Thread H36-pulled-up-after-being-held-up · Research | Racing Results
Threads · story · H36-pulled-up-after-being-held-up
H36: pulled up after being held up
Hypotheses
60 (28 alive)
Depth
16
Best rung
Live
Open questions
38
Last activity
17 Oct, 05:13
Where it stands: H-ade7c9c9 (Replicate: an unchanged mark after a 20+ length beating lifts close-fo) passed discovery as a niche: A/E 1.28 on 247 bets (p 0.146), same-race control 0.98.
The notebook
In order: what the lab saw, what it asked, what it proposed and why, what the tests said, and what an investigator concluded. Numbers are aggregates on the discovery data unless a step says otherwise.
If the debutant lift comes from the market over-reading one bad hurdle run, is it strongest when the horse finished close within its last three runs? testing
Does the parent held-up pull-up rule and its code-switch cuts (H-070b2026, H-7beeabeb) show the same GB-over-IRE split, or is that split specific to chase debutants? open
Is the heavily beaten debutant lift a winter-season effect that fades in summer jumping? It scored 1.45 before late April 2026 and 1.02 after. open
Is the broad heavy-beating lift carried by runners at 16/1 or bigger (1.174 on 3156), against the family pattern that longshots sit at or below par? testing
Does the heavy-beating family hold two separate lifts: big last fields under 16/1, and small last fields at 16/1 or bigger? open
Does a heavy beating in a big field mislead the market outside handicaps too, for example novice and maiden hurdlers? testing
Does the market's over-penalising of big-field heavy beatings also appear in Flat handicaps, where big fields are routine? testing
Is the big-field lift really a drop from a higher-class or festival handicap, rather than inflated beaten lengths? testing
Is the big-field heavy-beating lift specific to handicaps because big-field handicap beatings overstate the gap, while big-field non-handicaps are mostly no-hopers in maidens and novices? open
Does a class drop carry the heavy-beating lift independently of last-time field size in GB jumps handicaps? testing
Is the big-field heavy-beating lift jumps-only because jumps margins are inflated by easing and tiredness once beaten, which Flat races largely lack? testing
Is the longshot lift in the class-drop group (1.32 on 461 at 16/1 or bigger) carried by runners from small last fields, matching the split asked about in Q-a18717ff? open
Why does the never-involved lift vanish when last time's field had 17+ runners (1.005 on 461), while 11+ gives 1.33 and 24+ lengths in 16+ gives 1.437? open
Is the never-involved lift strongest when the heavy beating was a one-off, meaning the horse finished close in one of its other recent runs? testing
Why is the close-finish heavy-beating lift so much stronger in Ireland (2.11 on 153) than in GB (1.30 on 83)? open
Does the one-bad-run rebound replicate on the Flat once recent close form is required, moving the Flat lift out of the 16/1+ band? testing
Does the close-finish rebound hold below 16/1, or is it carried by a few longshot winners? open
Why is the ground/trip-change rebound so much stronger in IRE (1.53 on 298) than GB (1.18 on 619)? open
Is a change of conditions needed for the rebound, or does it work when the beating came on the same ground and at about the same trip? testing
How much do the four frozen jumps rules (recent winner, sub-16/1 winner, first run after win, ground/trip change) share the same runners, and does each keep a lift on the runners the others miss? testing
Why does IRE fail for the excuse-only rule (0.937) yet hold for the recent-winner rules (IRE 1.10)? Is Irish jumps pricing better at reading ground and trip excuses? open
Do recent winners priced 16/1 or bigger after a moderate (8 to 19 length) beating lose money consistently? (0.22 on 151 bets from 1 win) open
Is the non-handicap failure an IRE effect, where IRE non-handicap jumps are dominated by a few big yards whose beaten winners are correctly priced? open
Is the short-last-price niche's 16/1+ lift (6 winners from 44) a real market overreaction to a fancied flop, or noise from a handful of winners? open
Why did the price-band attack on H-40a72e24 count as held when it scored 1.004 on 182, below its stated kill line of 1.10? open
Is the unchanged-mark rebound strongest when the in-running comment offers a visible excuse (eased, pulled up, jumping error), or when the beating had no visible excuse and only the handicapper's leniency marks it? open
Does the unchanged-mark rebound after a heavy beating hold in Flat handicaps, where the plain close-form rebound failed (1.021 on 2055)? testing
Is the fall from the first period to the second (1.58 to 1.28 unchanged; 1.34 to 0.98 for non-winners) a seasonal effect of summer jumping, or the edge fading? open
Is an unchanged mark after the heavy beating often mechanical, because the beating came in a non-handicap or under the other code's rating, so the handicapper never judged the run? testing
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Promoted15 Oct, 23:10
H36: pulled up after being held upDiscovery → PROMISINGimport: tier A: strong candidate; best [14/1 or shorter] GREEN A/E 1.253 on 510 bets (broad AMBER)
because A horse held up and then pulled up was looked after once its chance had gone, so the run took little out of it. The market reads the P beside its name as a disaster. Twelve months of data only.
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND sp_decimal <= 15
because A held-up horse that was jumping better than most of its field and was still pulled up did not stop because of its jumping. Found by cutting the pulled-up system with measured jumping figures. Twelve months of data, 156 bets, and the two halves of the year disagree on how big the edge is. Unproven until its live record says otherwise.
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND sp_decimal <= 15 AND last_jump_index_pct >= 0.57
because A held-up horse that pulled up while jumping no worse than it normally does was looked after, not found out. The pulled-up horses to avoid are the good jumpers that jumped unusually badly: they won half as often as their prices implied. Found by cutting the pulled-up system with measured jumping figures. Twelve months of data and a small sample: 111 bets. Unproven until its live record says otherwise.
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND sp_decimal <= 15 AND usual_jump_index IS NOT NULL AND last_jump_index - usual_jump_index >= -0.15
H36-pulled-up-after-being-held-up (H36: pulled up after being held up) was GREEN in the earlier research: A/E 1.19 on 1451 bets (periods 1.07 / 1.41); now PROMISING.
Is the lift strongest when the last race was attritional, with many runners not finishing, so the PU says little about this horse?
When a large share of the field pulled up, the P is a property of the race (ground, pace, trip), not the horse, yet the form figure looks the same. Columns: last_race_dnf_share, last_field_size, last_pulled_up, last_run_style.
answeredpriority 0.50raised by O-309e6a2d
New question17 Oct, 01:33
Does the pulled-up lift depend on how quickly the horse is brought back, as it should if the run took nothing out of it?
If the horse was looked after, the trainer can bring it back quickly; a long absence after a PU points to injury or a problem the market is right to fear. Columns: days_since_last_run, last_pulled_up, last_run_style, sp_decimal. This tests the mechanism directly rather than re-cutting the pool by jumping figures.
answeredpriority 0.60raised by O-309e6a2d
New question17 Oct, 01:33
Is the A/E lift a spring/summer artefact, given the periods went 1.07 then 1.41?
The second period (from about April 2026) carries most of the lift. Smaller summer jumps fields and better ground could explain it rather than the PU story. Columns: race_date, field_size, going_group, race_type. Forward bets on the live system from October 2026 will partly answer it.
answeredpriority 0.35raised by O-309e6a2d
New question17 Oct, 01:33
Do pulled-up horses that led or raced prominently (and so likely stopped) fail to show the lift, confirming that run style is doing the work?
This is a negative control for the H36 mechanism. A PU after making the running usually means the horse stopped while trying. If those runners also score about 1.19, then the lift is a general PU discount and not 'held up and looked after'. Columns: last_pulled_up, last_run_style, sp_decimal.
To answer: Do pulled-up horses that led or raced prominently (and so likely stopped) fail to show the lift, confirming that run style is doing the work?
because If the market simply over-punishes any P, front-runners who pulled up should beat their price as well. If H36's mechanism is right, they stopped while trying, and their price is fair or too short.
and the market misses it because Under the rival, punters treat every P alike. Under H36, the front-runner's P really does reflect the horse failing, so it is priced correctly.
predicting negative, A/E 0.85 to 1.00 on 600+ bets (the H36 mechanism predicts no lift)
and it is wrong if A/E at or above 1.15, which would mean the held-up condition is not the cause
last_pulled_up = 1 AND last_run_style IN ('led', 'prominent')
To answer: Does the pulled-up lift depend on how quickly the horse is brought back, as it should if the run took nothing out of it?
because A horse that was held up, never got into the race and was pulled up while its chance was gone had a light, short race. The trainer can run it again within a few weeks. A long gap after a PU more often means the horse had a problem.
and the market misses it because The form line shows a P whatever the gap. Punters see a recent bad run and mark the horse down, while the trainer's quick return is a signal they overlook.
predicting positive, A/E 1.20 to 1.35 on 500+ bets, higher than the same pool with days_since_last_run > 35
and it is wrong if A/E below 1.05, or no higher than the long-gap complement
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND days_since_last_run IS NOT NULL AND days_since_last_run <= 35
To answer: Is the lift strongest when the last race was attritional, with many runners not finishing, so the PU says little about this horse?
because When a large share of a field does not finish (testing ground, a strong pace, a long trip), held-up runners are pulled up because the race fell apart in front of them. That says nothing about their ability or wellbeing.
and the market misses it because The form figure is the same P whether 1 or 6 runners pulled up. The context sits only in the full result, which most punters do not read.
predicting positive, A/E 1.25 to 1.45 on 300+ bets
and it is wrong if A/E below 1.10, or not above the low-DNF-share complement (last_race_dnf_share < 0.25)
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_dnf_share IS NOT NULL AND last_race_dnf_share >= 0.25
H36 is the root of a large branch of work. Held-up or midfield horses that were pulled up last time score A/E 1.19 on 1451 bets. The two periods disagree: 1.07 on 931 bets, then 1.41 on 520. The money is weak: BSP ROI is -5.7% and SP ROI is -34.6%, so the edge shows up in A/E but is mostly eaten by the price, especially at SP. The 14/1-or-shorter cut (GREEN, 1.253 on 510) went live on 2026-10-13 with no forward bets yet.
The jumping overlays (IQ60–67) have mostly reshuffled the same pool. Of those overlays:
- 'Jumped well' (1.35 on 156) and 'jumping as usual' (1.30 on 109) held up.
- 'Out of character bad jumping' collapsed (0.52 on 63).
- The 'poor jumper anyway' version failed its holdout outright (SYS1: 0.948 on 715).
So the cuts agree in direction: a pulled-up horse is forgivable when the PU was not caused by the horse failing. But every cut comes from the same twelve months and the same pool, and the holdout failure is a warning.
What the chain has not done is test the core cause from a different angle. Two consequences should hold if the cause is real:
- How soon the trainer brings the horse back (a horse that was looked after can return quickly).
- Whether the PU was a race-wide event rather than one horse failing.
There is also the obvious negative control: front-runners who stopped. Rival explanations already killed elsewhere (going change in H101, generic PU/F longshots in H-8841a2b0) should not be retried.
investigator v1
Tested17 Oct, 01:35
broad rule: A/E 1.03 on 777 bets, p = 0.46 (95%: 0.73 to 1.40), by period 0.97 · 1.22. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Killed17 Oct, 01:35
Test-mechanism: held-up PU horses brought back quickly beat their priceDiscovery → KILLEDfails: A/E 1.027, needs 1.05; p 0.4565, needs 0.05 or lower; periods [0.97, 1.22], each needs 1.0; A/E 1.027 against same-race control 1.0, needs 0.03 above (not a niche pass: A/E 1.027 below 1.05; does not beat the same-race control 1.0)
Tested17 Oct, 01:35
broad rule (niche): A/E 1.15 on 921 bets, p = 0.16 (95%: 0.88 to 1.46), by period 1.12 · 1.20. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:35
refinement: held-up only (purest form of 'race fell apart in front of them'): A/E 1.21 on 577 bets, p = 0.14 (95%: 0.85 to 1.67), by period 1.23 · 1.16. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:35
refinement: 14/1 or shorter today (story-family edges die at big prices): A/E 1.18 on 363 bets, p = 0.13 (95%: 0.89 to 1.55), by period 1.18 · 1.18. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:35
H-d61d2109 (Sharpen: held-up PU in an attritional race where many failed to finish) passed discovery as a niche: A/E 1.15 on 921 bets (p 0.157), same-race control 0.99.
result0.50reasoned about 17 Oct, 01:40from H-d61d2109
Tested17 Oct, 01:35
refinement: highly attritional last race (a third or more failed to finish): A/E 1.23 on 619 bets, p = 0.088 (95%: 0.91 to 1.64), by period 1.16 · 1.42. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:35
control: Primary control is the low-attrition complement: held-up/midfield runners pulled up last time in a race where u: A/E 1.28 on 530 bets, p = 0.11 (95%: 0.86 to 1.83), by period 0.94 · 1.69. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:36
H-d61d2109 (Sharpen: held-up PU in an attritional race where many failed to finish) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 01:40from H-d61d2109
Observed17 Oct, 01:36
H-a2870576 (Rival: pulled up after leading or racing prominently shows the same li) pointed the predicted way but failed the bar: A/E 0.92 on 727 bets (p 0.744), control 1.01.
near miss0.61reasoned about 17 Oct, 01:40from H-a2870576
Tested17 Oct, 01:36
broad rule: A/E 0.92 on 727 bets, p = 0.74 (95%: 0.67 to 1.22), by period 0.79 · 1.11. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:36
partition: Exclude last-run jumping errors (overlap with chase-error pool): A/E 1.09 on 691 bets, p = 0.30 (95%: 0.80 to 1.46), by period 1.07 · 1.14. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 01:40
Do lone held-up pull-ups (few other non-finishers in the race) beat their price more than pull-ups in attritional races?
The attritional cut came in below its parent (1.145 vs 1.198), implying about 1.3 for dnf_share below 0.25. A lone P looks equally bad in the form book, but it may mean a specific, fixable problem that the trainer then addresses. Columns: last_race_dnf_share, last_field_size, last_pulled_up, last_run_style, days_since_last_run.
answeredpriority 0.45raised by O-04d808cb, O-befbdb80
New question17 Oct, 01:40
Is the held-up PU lift concentrated in horses returning after a longer break (over 35 days), and does it rise with gap length or plateau?
The quick-return test failed at 1.027 on 777, so the implied long-gap remainder carries about 1.4. If a break after a held-up PU signals a fixed problem (wind op, freshening) that the market reads as injury, the lift should sit beyond 35 days and may persist to 90+ days. Columns: days_since_last_run, last_pulled_up, last_run_style, sp_decimal, first_headgear, race_date (season check).
answeredpriority 0.60raised by O-04d808cb, O-c2bb47dd
To answer: Do lone held-up pull-ups (few other non-finishers in the race) beat their price more than pull-ups in attritional races?
because When many runners pull up, the P tells you about the race, and the market discounts it anyway. When a held-up horse is one of few to pull up, it had a horse-specific, often transient or fixable issue and was nursed home rather than ground out. That issue is gone next time.
and the market misses it because A lone P looks worse to the market than a P in a race where half the field failed, so lone pull-ups drift further, but the underlying cause is often temporary.
predicting positive, A/E 1.20 to 1.40 on 350+ bets
and it is wrong if A/E below 1.10, or not above the attritional cut's 1.145
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_dnf_share IS NOT NULL AND last_race_dnf_share < 0.25
To answer: Is the held-up PU lift concentrated in horses returning after a longer break (over 35 days), and does it rise with gap length or plateau?
because A held-up horse pulled up because something specific was wrong (breathing, ground, a minor knock), and the trainer took it out of the race. A break of more than five weeks means the problem was addressed (a wind op, a freshening, waiting for ground). The horse returns sound, and its PU run took nothing out of it.
and the market misses it because The form line reads P followed by an absence, which punters and compilers read as a horse with a problem that is regressing. The fix is invisible in the form book unless first-time headgear or a wind-op note is flagged, and the absence makes the P look more ominous, not less.
predicting positive, A/E 1.25 to 1.50 on 450+ bets
and it is wrong if A/E below 1.10, or not clearly above the quick-return pool's 1.027, or the lift confined to one period
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND days_since_last_run IS NOT NULL AND days_since_last_run > 35
The thread now has three tests of H36's 'looked after, nothing taken out of it' story (held-up PU, A/E 1.198 on 1329). (1) The style control behaved as the mechanism predicts. Horses that pulled up after leading or racing prominently scored 0.916 on 727 against a same-race control of 1.01. The kill was formal: a lift-free population was never going to clear a significance bar in the negative direction. Its CI (0.67 to 1.22) is wide, so the 0.28 gap to the held-up pool is suggestive rather than proven. (2) The quick-return test came out the opposite way to its prediction: 1.027 on 777 at 35 days or fewer. Weighting roughly by bets, the remaining ~550 held-up PUs returning after more than 35 days must carry roughly 1.4. So whatever the lift is, it is not a horse 'brought straight back because nothing was wrong'. It looks more like a horse given a break or treated (wind, tie-back, freshening) and then underestimated on its return. (3) The attritional sharpen passed discovery as a niche at 1.145 on 921 and survived 5/5 attacks. But it sits below its parent, so the implied complement (lone PUs, dnf_share < 0.25, ~400 bets) scores about 1.3. That breaks its own falsification clause in spirit. Its frozen status reflects that it overlaps H36, not that the attritional mechanism is confirmed. Net belief: held-up style matters. The lift seems to sit where the PU was the horse's own event and it then had a break. Season remains a live confounder: period 2 (from April 2026) was 1.41, and long-gap returners cluster by season. Everything is still twelve months of discovery data, with 0 forward bets on the live system.
investigator v1
Answered17 Oct, 01:40
Is the lift strongest when the last race was attritional, with many runners not finishing, so the PU says little about this horse?
No. The attritional cut (dnf_share >= 0.25) scored 1.145 on 921, below the 1.198 parent, which implies lone pull-ups (dnf_share < 0.25) carry more of the lift, at roughly 1.3.
Does the pulled-up lift depend on how quickly the horse is brought back, as it should if the run took nothing out of it?
Yes, but the opposite way to the prediction: quick returns (35 days or fewer) scored only 1.027 on 777, so the lift implied for longer absences is roughly 1.4 on about 550 bets.
Do pulled-up horses that led or raced prominently (and so likely stopped) fail to show the lift, confirming that run style is doing the work?
Mostly yes. Led/prominent pulled-up horses showed no lift (A/E 0.916 on 727, control 1.01) against 1.198 for held-up ones, so run style is doing the work, though the negative side is not significant (CI 0.67 to 1.22).
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:41
refinement: break of 36 to 180 days (a freshening or a fix, not a long injury lay-off): A/E 1.30 on 632 bets, p = 0.044 (95%: 0.96 to 1.72), by period 1.22 · 1.41. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:41
refinement: no first-time headgear (the fix is invisible in the form book): A/E 1.38 on 580 bets, p = 0.020 (95%: 1.01 to 1.82), by period 1.25 · 1.54. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:41
H-6eae1245 (Test-mechanism: held-up PU horses returning after a break of more than) passed discovery: A/E 1.33 on 674 bets (p 0.024), same-race control 0.98.
result0.60reasoned about 17 Oct, 01:43from H-6eae1245
Tested17 Oct, 01:41
control: Main control: held-up/midfield pull-ups brought back within 35 days. The earlier finding puts this group at abo: A/E 1.03 on 777 bets, p = 0.46 (95%: 0.73 to 1.40), by period 0.97 · 1.22. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:42
control: Control: led/prominent pull-ups returning after more than 35 days: A/E 0.92 on 362 bets, p = 0.69 (95%: 0.60 to 1.35), by period 0.58 · 1.35. fail
H-6c3baaba (Rival: the held-up PU lift sits in lone pull-ups, not attritional race) pointed the predicted way but failed the bar: A/E 1.28 on 530 bets (p 0.106), control 0.98.
near miss0.70reasoned about 17 Oct, 01:43from H-6c3baaba
Observed17 Oct, 01:42
H-6c3baaba (Rival: the held-up PU lift sits in lone pull-ups, not attritional race): the period 'after 2026-04-27' carries the effect (A/E 1.69) while the rest score 2025-10-01 to 2026-04-27 0.94.
concentration0.55reasoned about 17 Oct, 01:43from H-6c3baaba
Tested17 Oct, 01:42
broad rule: A/E 1.28 on 530 bets, p = 0.11 (95%: 0.86 to 1.83), by period 0.94 · 1.69. fail
Is the held-up PU lift (and its long-gap concentration) really a change of ground between a winter pull-up and a spring or summer return, rather than the horse being freshened?
The long-gap cut passed discovery but broke under attack. The lone-PU lift is entirely after April 2026, and the parent's periods were 1.07 then 1.41. A horse pulled up on soft winter ground that returns after a break on better ground would explain all three facts at once. Columns: last_going_group, going_group, days_since_last_run, last_pulled_up, last_run_style.
answeredpriority 0.55raised by O-10828b53, O-abea699b, O-8033cc46
New question17 Oct, 01:43
Do smaller fields in the summer jumps season carry the late-period lift for held-up pull-ups?
The late-period concentration (1.69 against 0.94) needs a race-level explanation before any PU story is trusted. Smaller summer fields make win chances less spread out and SP less efficient. Columns: field_size, last_field_size, race_type, region.
answeredpriority 0.40raised by O-abea699b, O-435140b6
To answer: Is the held-up PU lift (and its long-gap concentration) really a change of ground between a winter pull-up and a spring or summer return, rather than the horse being freshened?
because A held-up horse that pulled up was often beaten by the ground (bogged down in testing going) rather than by ability or injury. Put it back on a different surface and the cause has gone. A long gap and a spring return are mostly the same horses: pulled up in winter mud, back on better ground.
and the market misses it because The form book shows a P and nothing about why. Punters price the P as a failure of the horse, not of that day's ground, and SP does not discount the change of conditions.
predicting positive, A/E 1.25 to 1.45 on 450+ bets, clearly above the same-going complement
and it is wrong if A/E below 1.10, or the same-going complement scores as high, which would leave freshening as the explanation
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_going_group IS NOT NULL AND going_group IS NOT NULL AND going_group <> last_going_group
To answer: Is the A/E lift a spring/summer artefact, given the periods went 1.07 then 1.41?
because Summer jumps cards have small fields, where a held-up horse that pulled up faces few rivals and is easier to place. The lift may be field size, not anything the horse kept back.
and the market misses it because In small fields the bookmakers' overround is spread across fewer runners, and the P makes the horse an easy one to lay. The price may be too long simply because there are few genuine alternatives.
predicting positive, A/E 1.30+ on 350+ bets if the field-size rival holds
and it is wrong if A/E at or below the parent's 1.198 in small fields, which would rule out field size as the carrier of the lift
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND field_size <= 9
The chain keeps landing on the same pool, held-up or midfield horses that pulled up last time (A/E 1.198 on 1329), and every split of it falls short on its own. Splitting by days since the last run gave the clearest contrast. The quick-return half scored 1.027 on 777. The long-gap half (over 35 days) scored 1.335 on 674 (p 0.024, BSP ROI +6.6%), then 2 of 5 falsification attacks broke it. Splitting by how many runners failed to finish in the last race gave a weaker contrast. The lone-PU half scored 1.28 on 530 (p 0.106) and the attritional half 1.145 on 921. Their intervals overlap heavily, so the lab cannot say which kind of race carries the lift.
The lone-PU result also lands entirely after 27 April 2026 (1.69, against 0.94 before). The parent's periods went 1.07 then 1.41. So the period skew now shows up in two separate cuts, and the seasonal worry grows rather than shrinks. Runners at 16/1 or bigger stay at or below par in both new cuts (0.775 and 1.045), so any lift is in the shorter-priced runners the live system already restricts to.
Current belief: the long-gap split is the best lead, but it is mixed up with the period. Horses pulled up in winter and brought back after a break run in spring and summer, on different ground and in smaller fields. The next tests should separate 'freshened after a PU' from 'summer conditions', not re-cut the pool again.
investigator v1
Answered17 Oct, 01:43
Do lone held-up pull-ups (few other non-finishers in the race) beat their price more than pull-ups in attritional races?
Not reliably. The lone-PU cut scored 1.28 on 530 (p 0.106) against the attritional cut's 1.145 on 921, the intervals overlap, and the lone-PU lift sits wholly after April 2026 (0.94 before).
Is the held-up PU lift concentrated in horses returning after a longer break (over 35 days), and does it rise with gap length or plateau?
In discovery, yes: over 35 days scored 1.335 on 674 against 1.027 on 777 for a quick return. But 2 of 5 attacks broke it. Whether the lift rises with gap length or levels off is still untested.
H-995c6f24 (Rival: held-up PU lift comes from a change of ground since the pull-up) passed discovery: A/E 1.3 on 940 bets (p 0.02), same-race control 0.98.
result0.60reasoned about 17 Oct, 01:48from H-995c6f24
Tested17 Oct, 01:46
broad rule: A/E 1.30 on 940 bets, p = 0.020 (95%: 1.01 to 1.64), by period 1.17 · 1.63. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:46
control: Main comparison is the same-going complement: held-up or midfield pull-ups returning on the same going group as: A/E 0.93 on 510 bets, p = 0.66 (95%: 0.58 to 1.41), by period 0.77 · 1.09. info
H-d084cd23 (Rival: held-up PU lift is a small-field summer artefact): the period 'after 2026-04-27' carries the effect (A/E 1.38) while the rest score 2025-10-01 to 2026-04-27 0.95.
concentration0.55reasoned about 17 Oct, 01:48from H-d084cd23
Observed17 Oct, 01:46
H-d084cd23 (Rival: held-up PU lift is a small-field summer artefact) pointed the predicted way but failed the bar: A/E 1.09 on 595 bets (p 0.282), control 0.99.
near miss0.62reasoned about 17 Oct, 01:48from H-d084cd23
Tested17 Oct, 01:46
broad rule: A/E 1.09 on 595 bets, p = 0.28 (95%: 0.82 to 1.42), by period 0.95 · 1.38. fail
Is the changed-going lift directional: strongest when the pull-up came on soft or heavy and today is on better ground, and absent when the move is towards softer ground?
The ground mechanism (the horse was bogged down, so the cause has gone) predicts direction. A lift for any change, whichever way, would point to a general 'different conditions' or season effect instead. H101 found no edge for soft-to-good across all pull-ups (1.02 on 3100), so the held-up condition must do the work. Columns: last_going_group, going_group, last_run_style, race_type.
answeredpriority 0.50raised by O-67404c8b, O-180aadda
New question17 Oct, 01:48
Why does the same-going complement also rise in the late period (0.77 then 1.09), and is there a season-wide drift in how held-up pull-ups are priced after April?
Both the changed and the same-going pools are stronger after 27 April. This points to a seasonal component that runs alongside the ground effect. Columns: race_date, going_group, field_size, race_class, race_type.
openpriority 0.30raised by O-67404c8b, O-180aadda
New question17 Oct, 01:48
Does the changed-going lift for held-up pull-ups survive among horses brought back within 35 days, where freshening cannot be the cause?
The changed-going cut (1.296 on 940) and the long-gap cut (1.335 on 674) probably overlap heavily. The quick-return pool as a whole was flat (1.027). If changed going lifts quick returners too, ground is the cause rather than rest. If it does not, freshening and season remain. Columns: days_since_last_run, going_group, last_going_group, last_pulled_up, last_run_style.
To answer: Is the changed-going lift directional: strongest when the pull-up came on soft or heavy and today is on better ground, and absent when the move is towards softer ground?
because A held-up horse pulled up on soft or heavy ground was bogged down and looked after once beaten. On sounder ground the cause has gone and the run took little out of it. The direction of the change is where the mechanism says the effect should be strongest.
and the market misses it because The P stands alone in the form figures, and punters read it as the horse failing. The going of the pull-up race is a step removed from the price. H101 shows the generic version (any pull-up) is priced, so any edge needs the held-up condition, which says the horse was looked after rather than stopping exhausted.
predicting positive, A/E 1.30 to 1.50 on 250+ bets
and it is wrong if A/E below 1.15, or not above the reverse-direction cut (better ground to softer), which would mean direction does not matter
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_going_group IN ('soft', 'heavy') AND going_group IS NOT NULL AND going_group NOT IN ('soft', 'heavy')
To answer: Is the changed-going lift directional: strongest when the pull-up came on soft or heavy and today is on better ground, and absent when the move is towards softer ground?
because Rival: connections switch conditions after a held-up pull-up for many reasons (a new trip, a new course, a target race). Any change, not escaping testing ground, would be what carries the lift. If so, moves from better ground to soft or heavy should score as well as the reverse.
and the market misses it because Whichever way the ground moved, the bare P is priced as a failure, so a general change-of-conditions effect would go unpriced too.
predicting positive, A/E 1.20+ on 200+ bets if the rival holds; the ground mechanism predicts about 1.0
and it is wrong if A/E below 1.10, or clearly below the mud-to-better cut. Either rules the rival out and supports the directional ground story.
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_going_group IS NOT NULL AND last_going_group NOT IN ('soft', 'heavy') AND going_group IN ('soft', 'heavy')
To answer: Does the changed-going lift for held-up pull-ups survive among horses brought back within 35 days, where freshening cannot be the cause?
because If the pull-up was caused by that day's ground and not by fatigue or lack of fitness, then a quick return on different ground removes the cause without any rest being needed. The lift should then appear in quick returners too.
and the market misses it because The form shows a recent P. A quick turnaround looks like the yard chasing a horse that has just failed, so the market lays it. The change of going is not priced as removing the cause.
predicting positive, A/E 1.15 to 1.35 on 300+ bets, clearly above the quick-return same-going pool
and it is wrong if A/E below 1.08, or no better than held-up pull-ups returning within 35 days on the same going. Either would leave freshening, not ground, as the cause.
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_going_group IS NOT NULL AND going_group IS NOT NULL AND going_group <> last_going_group AND days_since_last_run IS NOT NULL AND days_since_last_run <= 35
The chain now points at ground more than at freshening or field size. The parent held-up/midfield PU pool scored 1.198 on 1329. Splitting by gap put the lift in returns after more than 35 days (1.335 on 674, against 1.027 on 777 within 35 days), but that long-gap cut broke under 2 of 5 attacks. The ground rival (H-995c6f24) gave the cleanest split the thread has had. Horses running on a different going group from their pull-up scored 1.296 on 940 (p 0.02). Both periods were above par (1.17 on 635 up to late April, 1.63 on 305 after), and the same-race control was 0.98. The same-going complement scored 0.93 on 510, a gap of about 0.37. At 16/1 or shorter the changed-going cut scored 1.349 on 388 (BSP ROI +37%). The small-field rival (H-d084cd23) failed: 1.09 on 595, below the parent, with 0.36 on 294 at 16/1 or bigger. Field size does not carry the lift, and by arithmetic the 10+ runner remainder must sit above the parent. Two caveats keep this from being credited. First, a change of going travels with a longer gap and with season, so ground has not yet been separated from freshening. Second, the general idea of any pulled-up horse moving from soft to good (H101) was rejected at 1.02 on 3100. If ground matters, it does so only for horses that were looked after (held up), which is coherent with the original H36 story but still needs proving. In every split, the 16/1-or-bigger runners stay at or below par.
investigator v1
Answered17 Oct, 01:48
Is the A/E lift a spring/summer artefact, given the periods went 1.07 then 1.41?
Partly. The late skew is not field size (the small-field cut failed), and the changed-going cut beat par before late April too (1.17 on 635). The lift is not purely a spring/summer artefact, though the later period is still stronger (1.63).
Do smaller fields in the summer jumps season carry the late-period lift for held-up pull-ups?
No. Held-up pull-ups in fields of 9 or fewer scored 1.09 on 595, below the parent's 1.198, and 0.36 on 294 at 16/1 or bigger. Small summer fields do not carry the lift, so the larger fields must.
Tested17 Oct, 01:53
partition: Short gap band (35 days or less), 16/1 or shorter: A/E 1.11 on 171 bets, p = 0.34 (95%: 0.70 to 1.66), by period 1.07. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:53
partition: Hurdles only, 16/1 or shorter (ground cause without the chase-fatigue alternative): A/E 1.35 on 193 bets, p = 0.060 (95%: 0.92 to 1.90), by period 1.04 · 2.20. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:53
H-995c6f24 (Rival: held-up PU lift comes from a change of ground since the pull-up) passed discovery but was falsified: 2 of 5 attacks broke it.
near miss0.50reasoned about 17 Oct, 01:54from H-995c6f24
Tested17 Oct, 01:53
control: Control: held-up and well beaten but finished (no P), changed going, 16/1 or shorter: A/E 1.00 on 2,247 bets, p = 0.51 (95%: 0.88 to 1.13), by period 1.07 · 0.89. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:53
partition: Led or prominent pull-ups on changed going, 16/1 or shorter: A/E 0.94 on 243 bets, p = 0.66 (95%: 0.62 to 1.35), by period 0.87 · 1.09. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:53
H-dc628b02 (Test-mechanism: changed going lifts held-up pull-ups even on a quick r) passed discovery as a niche: A/E 1.09 on 449 bets (p 0.346), same-race control 1.0.
result0.50reasoned about 17 Oct, 01:54from H-dc628b02
Tested17 Oct, 01:53
refinement: shorter than 16/1: A/E 1.16 on 150 bets, p = 0.27 (95%: 0.73 to 1.74), by period 1.12. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:53
control: The key comparison is held-up or midfield pull-ups back within 35 days on the SAME going group as the pull-up. : A/E 0.90 on 327 bets, p = 0.68 (95%: 0.47 to 1.57), by period 0.80 · 1.03. info
H-658c845a (Sharpen: held-up pull-up in the mud, back on better ground) pointed the predicted way but failed the bar: A/E 1.05 on 377 bets (p 0.439), control 1.01.
near miss0.54reasoned about 17 Oct, 01:54from H-658c845a
Observed17 Oct, 01:53
H-658c845a (Sharpen: held-up pull-up in the mud, back on better ground): the period 'after 2026-04-27' carries the effect (A/E 1.56) while the rest score 2025-10-01 to 2026-04-27 0.87.
concentration0.55reasoned about 17 Oct, 01:54from H-658c845a
Tested17 Oct, 01:53
broad rule: A/E 1.05 on 377 bets, p = 0.44 (95%: 0.66 to 1.57), by period 0.87 · 1.56. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Killed17 Oct, 01:53
Sharpen: held-up pull-up in the mud, back on better groundDiscovery → KILLEDfails: A/E 1.048, needs 1.15; p 0.4388, needs 0.05 or lower; periods [0.87, 1.56], each needs 1.0 (not a niche pass: A/E 1.048 below 1.15; a period points the other way [0.87, 1.56])
Tested17 Oct, 01:53
control: Same-going quick-return control (no going change): A/E 0.90 on 327 bets, p = 0.68 (95%: 0.47 to 1.57), by period 0.80 · 1.03. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:53
H-bd36b503 (Rival: held-up pull-ups moving onto softer ground show the same lift) was predicted positive but came out the other way: A/E 0.92 on 149 bets (p 0.511 for the opposite direction).
surprise0.83reasoned about 17 Oct, 01:54from H-bd36b503
Observed17 Oct, 01:53
H-dc628b02 (Test-mechanism: changed going lifts held-up pull-ups even on a quick r) passed discovery but was falsified: driven by one price band (up to 14/1): the rest score 0.97; 2 of 5 attacks broke it.
near miss0.50reasoned about 17 Oct, 01:54from H-dc628b02
Tested17 Oct, 01:53
broad rule: A/E 0.92 on 149 bets, p = 0.63 (95%: 0.37 to 1.90), by period 1.01. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Killed17 Oct, 01:53
Rival: held-up pull-ups moving onto softer ground show the same liftDiscovery → KILLEDfails: 149 bets, needs 150; A/E 0.922, needs 1.1; p 0.6347, needs 0.05 or lower; periods [1.01], each needs 1.0; A/E 0.922 against same-race control 1.003, needs 0.03 above (not a niche pass: A/E 0.922 below 1.1; p 0.6347088871154392 above 0.5; does not beat the same-race control 1.003)
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 01:54
Which going transitions make up the roughly 400 changed-going held-up pull-ups that are neither mud-to-better nor better-to-softer, and do they include a change of surface or race code (jumps to Flat or AW)?
Changed going overall scored 1.296 on 940, but the two directional cuts total only 526 runners and score 1.05 and 0.92. The lift must sit in the residual. Scanner should cross-tab going_group x last_going_group, race_type x last_race_type, and last_surface, restricted to last_pulled_up = 1 AND last_run_style IN ('held_up','midfield').
answeredpriority 0.60raised by O-2975deac, O-00e05ce8, O-fcb8f2fd
New question17 Oct, 01:54
Is the after-April-2026 concentration a feature of every held-up pull-up cut, and does it go with a change in race mix (race_type, race_class, is_handicap) rather than with anything about the horse?
Mud to better ground went 0.87 then 1.56, and earlier cuts showed the same late skew. A season-wide shift would mean every sub-cut is really measuring the calendar. Columns: race_date, race_type, race_class, is_handicap, field_size, region.
To answer: Which going transitions make up the roughly 400 changed-going held-up pull-ups that are neither mud-to-better nor better-to-softer, and do they include a change of surface or race code (jumps to Flat or AW)?
because After a pull-up, connections often switch a held-up horse to a code that suits it better (fences, or the Flat or AW in summer). A change of going group comes with that switch, especially onto AW or summer Flat ground. The pull-up told us nothing about the horse in the new code.
and the market misses it because The form line shows a P from a different discipline, and punters weigh it as if it were a failure in today's code.
predicting positive, A/E 1.20 to 1.40 on 300+ bets, above the same-code complement
and it is wrong if A/E below 1.10, or the same-code held-up pull-ups score as high
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type IS NOT NULL AND race_type <> last_race_type
The chain now points away from ground as the cause. Held-up or midfield pull-ups (H36) scored 1.198 on 1329. The changed-going cut scored 1.296 on 940 against 0.93 on 510 for same going, which looked like the cause. But the two directional cuts that make up that split do not carry it. Moving from mud to better ground scored only 1.05 on 377 (control 1.01; 0.87 before 27 April, then 1.56). Moving onto softer ground scored 0.92 on 149, the opposite of what the rival predicted, and the ground story said it should be flat. Changed going among quick returners scored 1.09 on 449, about the same as the flat quick-return pool (1.027). It was killed because only the up-to-14/1 band carried it (the rest scored 0.97). The changed-going cut and the long-gap cut (1.335 on 674) each broke on 2 of 5 attacks. So the 'bogged down in winter mud, back on spring ground' story fails. About 400 changed-going runners fall in neither directional cut, and they must hold most of the lift. That residual has to be identified before ground is credited with anything. Two things recur in every cut: runners at 16/1 or bigger sit at or below par, and the effect sits in the period after April 2026. The thread's durable asset is still the parent's 14/1-or-shorter band (1.253 on 510) and the attritional-race cut in shadow (1.145 on 921, 5/5 attacks). Every narrowing of the mechanism since has been killed.
investigator v1
Answered17 Oct, 01:54
Is the held-up PU lift (and its long-gap concentration) really a change of ground between a winter pull-up and a spring or summer return, rather than the horse being freshened?
Neither cleanly. The winter-mud-to-better-ground story fails (1.05 on 377), and the freshening (long-gap) cut broke under attack, so neither ground nor freshening alone explains the held-up pull-up lift.
Is the changed-going lift directional: strongest when the pull-up came on soft or heavy and today is on better ground, and absent when the move is towards softer ground?
No. Mud to better ground scored only 1.05 on 377 (p 0.44), and moves onto softer ground scored 0.92 on 149. The directional ground move does not carry the changed-going lift of 1.296.
Does the changed-going lift for held-up pull-ups survive among horses brought back within 35 days, where freshening cannot be the cause?
Barely. Changed going among horses back within 35 days scored 1.09 on 449, close to the quick-return pool's 1.027. Only the up-to-14/1 band carried it, and it was killed (rest 0.97, 2 of 5 attacks broke it).
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:58
refinement: code switch on the same going group (separates code from going): A/E 1.42 on 99 bets, p = 0.21 (95%: 0.61 to 2.80), by period 0.88 · 1.79. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:58
H-070b2026 (Rival: held-up PU lift comes from a switch of race code since the pull) passed discovery as a niche: A/E 1.36 on 304 bets (p 0.092), same-race control 0.98.
result0.50reasoned about 17 Oct, 02:03from H-070b2026
Tested17 Oct, 01:58
control: Same-race other runners should be near par (about 0.95 to 1.02). The key comparison is the same-code complement: A/E 1.14 on 1,147 bets, p = 0.15 (95%: 0.89 to 1.44), by period 0.99 · 1.45. info
control: Control: same-code held-up/midfield pull-ups, 14/1 or shorter: A/E 1.17 on 403 bets, p = 0.14 (95%: 0.89 to 1.51), by period 1.02 · 1.48. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 01:59
partition: Partition: jumps-to-jumps switch only (Hurdle and Chase), no Flat or Bumper: A/E 1.42 on 226 bets, p = 0.078 (95%: 0.87 to 2.20), by period 1.46 · 1.36. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 01:59
H-070b2026 (Rival: held-up PU lift comes from a switch of race code since the pull) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 02:03from H-070b2026
Tested17 Oct, 01:59
partition: Partition: jumps to Flat switch only: A/E 1.33 on 64 bets, p = 0.39 (95%: 0.27 to 3.89). pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 02:03
Does any last-time non-completion lose its meaning once the horse switches code, or is the effect specific to a held-up pull-up?
raised with hypothesis 'Generalise: a fall or unseat followed by a switch of race code'
testingraised by O-9a6df59c, O-4aec4e49
New question17 Oct, 02:03
Within the 226 jumps-to-jumps switchers, is the lift in hurdle-to-chase (a first try over fences) or chase-to-hurdle (fences taken away), and does a jumping error in the pull-up matter?
The code switch carries a lift that holds in both periods (1.44, then 1.24), and jumps-to-jumps switches make up most of it (1.42 on 226). Which direction carries it would tell us whether the cause is new conditions or obstacles the horse could not cope with. Columns: last_race_type, race_type, chase_runs, hurdle_runs, last_jumping_error, jumping_errors_last3, sp_decimal.
answeredpriority 0.60raised by O-9a6df59c, O-4aec4e49
New question17 Oct, 02:03
Why does the late-period lift sit only in same-code held-up pull-ups (0.99, then 1.45) and not in code switchers, and does it follow summer going, race class or handicap mix?
Code switchers lack the late skew, so the calendar effect belongs to same-code runners. If it follows summer jumping on good or firm ground, or low-class handicaps, it is a race-mix effect and not about the horse. Columns: race_date, going_group, race_class, is_handicap, field_size, race_type.
openpriority 0.40raised by O-9a6df59c
New question17 Oct, 02:03
If the code-switch lift comes from the market reading a pull-up in another code as a failure, is it strongest when the horse has little or no form in today's code?
raised with hypothesis 'Test-mechanism: held-up pull-up switching to a code the horse has barely tried'
To answer: Does any last-time non-completion lose its meaning once the horse switches code, or is the effect specific to a held-up pull-up?
because A non-completion in one code says little about ability in another, especially chase falls followed by a drop back to hurdles, where the obstacle that caused the failure is removed.
and the market misses it because Form readers treat F or U as a black mark whatever the code, so the horse drifts in a discipline where the failure does not apply.
predicting positive, A/E 1.10 to 1.30 on 250+ bets
and it is wrong if A/E at or below 1.05, which would make the held-up pull-up the essential part
last_fell_or_unseated = 1 AND last_race_type IS NOT NULL AND race_type <> last_race_type
To answer: If the code-switch lift comes from the market reading a pull-up in another code as a failure, is it strongest when the horse has little or no form in today's code?
because A held-up horse that was pulled up and is now tried in a new code (first or second chase, hurdle or Flat run) has no form in today's discipline, and the last-time P is all the market has. Connections switch codes because they expect the new one to suit.
and the market misses it because The form line shows a recent P and nothing in today's code to set against it, so the price leans on the failure. Horses with plenty of runs in today's code are already rated on that form.
predicting positive, A/E 1.35 to 1.60 on 150+ bets, above the experienced-in-code switchers
and it is wrong if A/E below 1.15, or no higher than switchers with 3 or more runs in today's code
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type IS NOT NULL AND race_type <> last_race_type AND ((race_type = 'Chase' AND COALESCE(chase_runs, 0) <= 1) OR (race_type = 'Hurdle' AND COALESCE(hurdle_runs, 0) <= 1) OR (race_type = 'Flat' AND COALESCE(flat_runs, 0) <= 1))
To answer: Within the 226 jumps-to-jumps switchers, is the lift in hurdle-to-chase (a first try over fences) or chase-to-hurdle (fences taken away), and does a jumping error in the pull-up matter?
because After a pull-up, connections often change the type of race: into a novice or maiden race, or into a handicap off a dropped mark. Any change of conditions makes the last run less relevant, which matches the code-switch story.
and the market misses it because If the generic explanation is right, the market over-weights a P whenever conditions change, not only when the code changes.
predicting positive, A/E 1.20+ on 250+ bets if the rival holds
and it is wrong if A/E below 1.10 with the late-period skew of other same-code cuts, which would leave the code switch itself as the carrier
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND race_type = last_race_type AND last_is_handicap IS NOT NULL AND is_handicap IS NOT NULL AND is_handicap <> last_is_handicap
The thread now points to a better carrier than the going. Held-up or midfield pull-ups running today in a different race type scored A/E 1.36 on 304 (p 0.09, niche pass, control 0.98). The result held in both periods, 1.44 on 170 before 27 April 2026 and 1.24 on 134 after, and it survived 6 of 6 attacks. That makes it the first cut in this thread without the late-period skew.
The same-code complement tells a different story. It scored 1.14 on 1147, and that figure is entirely late: 0.99 before 27 April, 1.45 after. Same-code runners on changed going at 14/1 or shorter scored 1.30 on 277, also late (1.06, then 2.03). So the earlier changed-going lift (1.296 on 940) looks like two things mixed together: a stable code-switch effect, and a calendar effect in the same-code runners.
Hurdle-to-chase and chase-to-hurdle switches carry most of the effect (1.42 on 226). Jumps to Flat (1.33 on 64, 3 wins) is too thin to read. The lift needs the pull-up: code switchers in general score 1.007 on 3532. It is also weaker without the held-up style, at 1.19 on 476, which implies roughly 0.9 for led or prominent switchers. That fits the H36 mechanism: a held-up pull-up says little about the horse, and less again in another code.
As everywhere in this thread, outsiders are dead. Runners at 16/1 or bigger scored 0.55 on 197, against 1.58 on 107 at 14/1 or shorter. The code switch on the same going is too small to settle anything (8 wins).
investigator v1
Answered17 Oct, 02:03
Which going transitions make up the roughly 400 changed-going held-up pull-ups that are neither mud-to-better nor better-to-softer, and do they include a change of surface or race code (jumps to Flat or AW)?
Yes: about 300 held-up pull-ups changed race code, mostly hurdle to chase or back (226). They score 1.36, steady across periods, while same-code runners on changed going owe their 1.30 (at 14/1 or shorter) to the late period.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 02:09
H-658519da (Test-mechanism: held-up pull-up switching to a code the horse has bare) passed discovery as a niche: A/E 1.26 on 53 bets (p 0.394), same-race control 0.98.
result0.50reasoned about 17 Oct, 02:13from H-658519da
Observed17 Oct, 02:09
H-525d10d2 (Generalise: a fall or unseat followed by a switch of race code) was predicted positive but came out the other way: A/E 0.75 on 188 bets (p 0.211 for the opposite direction).
surprise0.90reasoned about 17 Oct, 02:13from H-525d10d2
Tested17 Oct, 02:09
broad rule: A/E 0.75 on 188 bets, p = 0.86 (95%: 0.38 to 1.35), by period 0.69 · 0.85. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:09
control: Two comparisons. (1) The lab's same-race control: the other runners in these races should sit near par (about 0: A/E 1.35 on 228 bets, p = 0.14 (95%: 0.79 to 2.16), by period 1.37 · 1.32. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Killed17 Oct, 02:09
Generalise: a fall or unseat followed by a switch of race codeDiscovery → KILLEDfails: A/E 0.752, needs 1.05; p 0.8622, needs 0.05 or lower; periods [0.69, 0.85], each needs 1.0; A/E 0.752 against same-race control 1.019, needs 0.03 above (not a niche pass: A/E 0.752 below 1.05; p 0.8621942564032626 above 0.5; does not beat the same-race control 1.019; a period points the other way [0.69, 0.85])
H-d105f080 (Rival: a change of handicap status, not of code, carries the switcher ): the period 'after 2026-04-27' carries the effect (A/E 2.92) while the rest score 2025-10-01 to 2026-04-27 0.82.
concentration0.55reasoned about 17 Oct, 02:13from H-d105f080
Observed17 Oct, 02:09
H-d105f080 (Rival: a change of handicap status, not of code, carries the switcher ) pointed the predicted way but failed the bar: A/E 1.41 on 112 bets (p 0.151), control 0.96.
near miss0.60reasoned about 17 Oct, 02:13from H-d105f080
Tested17 Oct, 02:09
broad rule: A/E 1.41 on 112 bets, p = 0.15 (95%: 0.73 to 2.47), by period 0.82 · 2.92. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:11
control: Control: barely-tried switchers who were NOT pulled up last time: A/E 1.02 on 1,203 bets, p = 0.44 (95%: 0.84 to 1.23), by period 0.91 · 1.08. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 02:11
H-658519da (Test-mechanism: held-up pull-up switching to a code the horse has bare) survived falsification: 3 of 3 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 02:13from H-658519da
Tested17 Oct, 02:11
control: Control: prominent or front-running pull-ups switching to a barely-tried code: A/E 0.00 on 29 bets, p = 1.0 (95%: 0.00 to 1.86). info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 02:13
Does every same-code held-up pull-up cut, including a change of handicap status, carry its lift only after 27 April 2026, and is that explained by the summer race mix (race_class, is_handicap, going_group)?
The handicap-status cut went 0.82 then 2.92, adding one more cut to the list with this calendar skew. The code-switch cut does not show it. A descriptive split by race_date crossed with race_class, is_handicap and going_group would show whether the late lift is a race-mix effect. Columns: race_date, race_class, is_handicap, last_is_handicap, going_group, race_type.
openpriority 0.40raised by O-1498078c, O-800b0627
New question17 Oct, 02:13
Is the code-switch lift specific to pull-ups that were held up, or does any last-time pull-up (including prominent or front-running ones) that switches code beat its price?
Fallers and unseaters who switch code score 0.75, so the switch alone is not the cause. If prominent pull-ups that switch also score around 1.3, the cause is the pull-up plus the switch rather than being held up. If they score at or below par, the 'looked after, form hidden' mechanism stands. Columns: last_pulled_up, last_run_style, last_race_type, race_type, sp_decimal.
answeredpriority 0.55raised by O-d9f14aca, O-4aec4e49
New question17 Oct, 02:13
Are last-time fallers and unseaters switched to a different code over-bet, and is it worst when chasers drop back to hurdles?
The opposite-way result (0.75 on 188, below control in both periods and both countries) suggests the market may credit connections' fix of switching code too much after a crash. It is not significant (p 0.21), so treat it as a lead. Columns: last_fell_or_unseated, last_race_type, race_type, sp_decimal, last_jumping_error.
To answer: Is the code-switch lift specific to pull-ups that were held up, or does any last-time pull-up (including prominent or front-running ones) that switches code beat its price?
because Rival view: the market over-penalises any pull-up, and a switch of code resets the form so that run counts for less. If so, prominent and front-running pull-ups that switch code should also beat their price. The held-up mechanism predicts they will not, because they were not looked after and their pull-up reflects real weakness.
and the market misses it because Under the rival, the P figure is marked down the same way whatever the run style, while a change of code makes it irrelevant.
predicting positive, A/E 1.20+ on 150+ bets if the rival holds
and it is wrong if A/E at or below 1.05, which would rule out the rival and support the held-up mechanism
last_pulled_up = 1 AND last_run_style IS NOT NULL AND last_run_style NOT IN ('held_up', 'midfield') AND last_race_type IS NOT NULL AND race_type <> last_race_type
To answer: Within the 226 jumps-to-jumps switchers, is the lift in hurdle-to-chase (a first try over fences) or chase-to-hurdle (fences taken away), and does a jumping error in the pull-up matter?
because A hurdler that was held up and pulled up was looked after once beaten, and the hurdle race may not have suited it. Fences and the chase test can bring out a horse bred for chasing. A chaser dropped back to hurdles after a crash is what the fallers cut covers, and that cut scored 0.75, which argues against 'fences taken away' as the carrier.
and the market misses it because The form figures show a P on top of a first run over fences. Punters discount a horse that is unproven over fences and pulled up last time. Nothing in the form book says the horse was given an easy time once held up.
predicting positive, A/E 1.30 to 1.50 on 100+ bets
and it is wrong if A/E below 1.10, or not above the chase-to-hurdle cut of held-up pull-ups
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Hurdle' AND race_type = 'Chase'
The held-up pull-up thread has narrowed to one lead: the race-code switch. H-070b2026 (1.359 on 304) is in shadow after holding all 6 attacks, and it is the only cut whose lift holds in both periods. This round tested where the switch lift comes from.
- **Fallers and unseaters:** the switch lift does not extend to them. They scored A/E 0.75 on 188 bets, below the same-race control of 1.02 in both periods (0.69 and 0.85), and both countries were below 0.8. This came out the opposite way to the prediction. So a switch of code is not a cure for any failed run. The lift seems to need a pull-up from a held-up position, where the horse was looked after and its form was hidden. A crash in public is different, and that horse may then be over-bet when its code is switched.
- **Handicap status:** changing handicap status within the same code (H-d105f080) scored 1.41 on 112 bets. But it was 0.82 before 27 April and 2.92 after, the same calendar pattern that killed the other same-code cuts. It is not a separate cause.
- **Barely-tried code (H-658519da):** it passed only as a niche, 1.26 on 53 bets. It went 0 from 34 at 16/1 or bigger, and its lift rests on 2 winners from 10 bets at up to 14/1. It held all 3 attacks but is far too thin to support a mechanism. The forward record will judge it.
What we believe now:
- The parent lift of about 1.2 is the late-season calendar effect plus the code-switch subset.
- The code switch is the only part not tied to the calendar, and it appears to be specific to held-up pull-ups.
- Which way the switch goes (hurdles to fences, or fences to hurdles) is still unknown. The faller result makes "fences taken away" less likely as a general story.
investigator v1
Observed17 Oct, 02:15
H-fbabeb50 (Sharpen: held-up hurdle pull-ups sent chasing next time) passed discovery as a niche: A/E 1.5 on 76 bets (p 0.17), same-race control 0.96.
result0.50reasoned about 17 Oct, 02:31from H-fbabeb50
Tested17 Oct, 02:15
broad rule (niche): A/E 1.50 on 76 bets, p = 0.17 (95%: 0.65 to 2.96), by period 1.77. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:15
control: Two comparisons. (1) Same-race rivals: the other runners in these chases should sit near A/E 1.0. Their score s: A/E 1.38 on 150 bets, p = 0.17 (95%: 0.71 to 2.40), by period 1.28 · 1.55. info
H-f0aa49df (Rival: any pull-up switching code beats its price, held-up or not) was predicted positive but came out the other way: A/E 0.87 on 161 bets (p 0.414 for the opposite direction).
surprise0.88reasoned about 17 Oct, 02:31from H-f0aa49df
Tested17 Oct, 02:15
broad rule: A/E 0.87 on 161 bets, p = 0.71 (95%: 0.40 to 1.65), by period 0.56 · 1.20. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Killed17 Oct, 02:15
Rival: any pull-up switching code beats its price, held-up or notDiscovery → KILLEDfails: A/E 0.869, needs 1.05; p 0.706, needs 0.05 or lower; periods [0.56, 1.2], each needs 1.0; A/E 0.869 against same-race control 1.011, needs 0.03 above (not a niche pass: A/E 0.869 below 1.05; p 0.7060174416015405 above 0.5; does not beat the same-race control 1.011; a period points the other way [0.56, 1.2])
Tested17 Oct, 02:15
control: Control: held-up hurdle pull-ups staying over hurdles: A/E 1.24 on 717 bets, p = 0.10 (95%: 0.89 to 1.69), by period 0.92 · 1.82. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 02:15
H-fbabeb50 (Sharpen: held-up hurdle pull-ups sent chasing next time) survived falsification: 4 of 4 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 02:31from H-fbabeb50
Tested17 Oct, 02:15
control: Control: hurdle pull-ups that raced prominently or led, sent chasing: A/E 1.04 on 53 bets, p = 0.54 (95%: 0.28 to 2.65). pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:15
control: Control: held-up hurdlers that completed last time, sent chasing: A/E 1.05 on 759 bets, p = 0.32 (95%: 0.85 to 1.29), by period 0.99 · 1.10. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:15
partition: Partition: handicap chases only (removes novice-chase pricing effects): A/E 1.10 on 65 bets, p = 0.48 (95%: 0.36 to 2.56), by period 1.33. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 02:31
Is the hurdle-to-chase lift about the pull-up hiding a horse's ability, or does any heavily beaten last hurdle run before a first chase beat its price, pulled up or not?
H-fbabeb50 scored 1.50 on 76 against a same-race control of 0.96. Chasing debutants are often under-rated after poor hurdle form whatever the reason for it. If debutants beaten 20 lengths or more over hurdles also score around 1.3, the pull-up is not the essential ingredient. Columns: last_race_type, race_type, chase_runs, last_beaten_lengths, last_pulled_up, last_run_style.
answeredpriority 0.50raised by O-2b845062, O-cd625e74
To answer: Is the hurdle-to-chase lift about the pull-up hiding a horse's ability, or does any heavily beaten last hurdle run before a first chase beat its price, pulled up or not?
because Big chasing types often run poorly over hurdles because they are too slow or too big for that game, then improve for fences. Any heavy defeat would hide that, not only a pull-up.
and the market misses it because Punters price the debut off the recent hurdle figures, and a horse beaten 20 lengths or more looks out of form even though the fences are the point of the switch.
predicting positive, A/E 1.15 to 1.30 on 300+ bets if the rival holds
and it is wrong if A/E at or below 1.05 or not above the same-race control. That would leave the held-up pull-up as the essential part of H-fbabeb50.
last_race_type = 'Hurdle' AND race_type = 'Chase' AND COALESCE(chase_runs, 0) = 0 AND COALESCE(last_pulled_up, 0) = 0 AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20
To answer: Within the 226 jumps-to-jumps switchers, is the lift in hurdle-to-chase (a first try over fences) or chase-to-hurdle (fences taken away), and does a jumping error in the pull-up matter?
because If the cause is a pull-up that hides ability (the horse was looked after, never asked a question) and today's code change resets the market's reading, then chase-to-hurdle switchers should lift as well. If the cause is new conditions suiting the horse (a chasing type finally over fences), the lift should be weak or absent here.
and the market misses it because The P in the form figures and a code change both push the price out. A chaser dropped back to hurdles after pulling up looks like a horse in decline, not one that was simply looked after.
predicting positive, A/E 1.20 to 1.40 on 120+ bets if the hidden-ability mechanism holds
and it is wrong if A/E below 1.05 or not above the same-race control. That would put the switch lift in the hurdle-to-chase direction, meaning new conditions rather than form that misled.
last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Chase' AND race_type = 'Hurdle'
The code-switch lift belongs to pull-ups that were held up. Held-up or midfield pull-ups that switch code scored A/E 1.36 on 304 bets and survived 6 of 6 attacks. Pull-ups that raced prominently or led and then switched code scored 0.87 on 161 (O-cd625e74), below their same-race control of 1.01, with periods of 0.56 and 1.20. Fallers and unseaters who switched scored 0.75. So the switch alone does nothing. The lift needs a pull-up in which the horse was never asked a question. A prominent pull-up usually means the horse stopped after a real effort, and that is honest bad form. A held-up pull-up was looked after, so the form book overstates how bad it was. In the prominent group the only sign of life is 2 winners at 16/1 or bigger (1.22 on 86), which is noise. Within the switchers, hurdle-to-chase held-up pull-ups scored 1.50 on 76 (O-2b845062). The CI runs from 0.65 to 2.96, p is 0.17 and the same-race control is 0.96. It passed as a niche, held all 4 attacks and is frozen for the forward record (O-f1f004ec). Its point estimate is above the switch parent's 1.36, but it is far too thin to say the fences direction carries the lift. The chase-to-hurdle half of the jumps switch has not been measured, and that is the missing comparison. Elsewhere in the thread, every going-based explanation is dead. Cuts within the same code keep showing a lift only after late April 2026, which looks like a calendar artefact rather than a cause.
investigator v1
Answered17 Oct, 02:31
Is the code-switch lift specific to pull-ups that were held up, or does any last-time pull-up (including prominent or front-running ones) that switches code beat its price?
The lift is specific to held-up pull-ups. Prominent or front-running pull-ups that switch code scored A/E 0.87 on 161 bets, against a same-race control of 1.01, while held-up switchers scored 1.36 on 304.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:35
control: The same-race control (other runners in the same hurdle races) should sit near A/E 1.0, which shows the race ty: A/E 1.08 on 404 bets, p = 0.35 (95%: 0.73 to 1.55), by period 1.09 · 1.08. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 02:35
H-7beeabeb (Test-mechanism: held-up chase pull-ups dropped back to hurdles) passed discovery as a niche: A/E 1.38 on 150 bets (p 0.17), same-race control 0.98.
result0.50reasoned about 17 Oct, 02:39from H-7beeabeb
Observed17 Oct, 02:35
H-448cb451 (Rival: chase debutants after any heavy hurdle beating beat their price): the country 'GB' carries the effect (A/E 1.43) while the rest score IRE 0.91.
concentration0.55reasoned about 17 Oct, 02:39from H-448cb451
Observed17 Oct, 02:35
H-448cb451 (Rival: chase debutants after any heavy hurdle beating beat their price) passed discovery as a niche: A/E 1.24 on 335 bets (p 0.116), same-race control 0.97.
result0.50reasoned about 17 Oct, 02:39from H-448cb451
Tested17 Oct, 02:35
broad rule (niche): A/E 1.24 on 335 bets, p = 0.12 (95%: 0.87 to 1.71), by period 1.53 · 0.99. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:35
refinement: held up or midfield last time (does the H-fbabeb50 run style still matter among finishers?): A/E 1.08 on 192 bets, p = 0.40 (95%: 0.65 to 1.68), by period 0.95 · 1.19. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:35
control: The lab's same-race control (the other runners in these novice and handicap chases) should sit near 1.0. If the: A/E 0.91 on 229 bets, p = 0.75 (95%: 0.66 to 1.22), by period 0.94 · 0.89. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 02:39
Is the chase-debutant lift after a heavy hurdle beating carried by yards that are good at chase debuts, so that it is really the published yard effect (H104, SYS-chase-debut-from-a-good-yard) and not a misread of form?
H-448cb451 passed as a niche at 1.24 on 335, and good chase-debut yards already show 1.18. If the debutants come mostly from those yards, the 'bad hurdle form misread' story adds nothing. A scan would read tx_chase_debut_runs, tx_chase_debut_wins, tx_chase_debut_exp, chase_runs, last_beaten_lengths and last_race_type.
answeredpriority 0.60raised by O-7628ef4c, O-69d892e9
New question17 Oct, 02:39
Is the GB-only concentration of the debutant lift (GB 1.43, IRE 0.91) a property of the race mix, such as GB novice handicap chases or race class, rather than the country?
The country split came from a scan. GB debutants run more often in novice handicap chases, where the handicapper's mark after bad hurdle form may be lenient. A descriptive split of region by is_handicap, is_novice, race_class and last_official_rating would show whether a race property explains it before anyone cuts the rule by country.
openpriority 0.40raised by O-69d892e9
New question17 Oct, 02:39
Across every code-switch cut in this thread, does the lift sit entirely under 16/1, and is that more than the general longshot bias of jumps racing?
Chase-to-hurdle pull-ups had 0 winners from 94 bets at 16/1 or bigger; H-070b2026 scored 2.22 under 14/1 against 0.55 at 16/1 or bigger; the debutants scored 1.03 at 16/1 or bigger. Comparing the same price bands for all jumps runners with a bad last run (sp_decimal, last_beaten_lengths, last_pulled_up) would show whether this is the mechanism or just the base rate.
openpriority 0.45raised by O-3bed9331, O-7628ef4c
New question17 Oct, 02:39
If the debutant lift comes from the market over-reading one bad hurdle run, is it strongest when the horse finished close within its last three runs?
raised with hypothesis 'Test-mechanism: chase debutants beaten heavily last time over hurdles but close '
To answer: Is the chase-debutant lift after a heavy hurdle beating carried by yards that are good at chase debuts, so that it is really the published yard effect (H104, SYS-chase-debut-from-a-good-yard) and not a misread of form?
because If the cause is the market over-reading bad hurdle form, the lift should appear whoever the trainer is. If it disappears outside the good chase-debut yards, the published yard effect is the real driver.
and the market misses it because Bad hurdle form puts the horse out of the betting regardless of the yard. If the lift survives in ordinary yards, the misread is about the form and not about the trainer's schooling.
predicting positive, A/E 1.15 to 1.30 on 200+ bets if the form-misread story holds
and it is wrong if A/E at or below 1.05 on 150+ bets, which would mean the yard carries the effect and the story adds nothing
last_race_type = 'Hurdle' AND race_type = 'Chase' AND COALESCE(chase_runs, 0) = 0 AND last_pulled_up = 0 AND last_beaten_lengths >= 20 AND (COALESCE(tx_chase_debut_runs, 0) < 15 OR tx_chase_debut_wins < 1.3 * tx_chase_debut_exp)
To answer: If the debutant lift comes from the market over-reading one bad hurdle run, is it strongest when the horse finished close within its last three runs?
because A horse that was beaten heavily last time but finished within a few lengths in one of the two runs before it has shown ability that the last run hid. The move to fences gives it new conditions in which that ability can come out again.
and the market misses it because The race card and punters lead with the last run, so a 20-length beating dominates the price, while the closer run two or three back is discounted. Fit, winning hurdlers (H124, 0.924) are priced correctly, so the error is specific to bad recent form.
predicting positive, A/E 1.30 to 1.50 on 150+ bets, above the debutants with no close recent run
and it is wrong if A/E below 1.15, or no higher than heavily beaten debutants whose best of the last three runs was also beaten more than 10 lengths
last_race_type = 'Hurdle' AND race_type = 'Chase' AND COALESCE(chase_runs, 0) = 0 AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6
The thread has narrowed to one claim: a held-up pull-up loses its meaning when the horse switches code (H-070b2026, 1.359 on 304, in shadow with no forward bets yet). These observations say the lift is not specific to the pull-up or to one direction of switch. First, chase debutants after a hurdle run beaten 20 lengths or more, with pull-ups excluded, scored 1.24 on 335 (p 0.116, same-race control 0.97, CI 0.87 to 1.71). That is a wider pool than the pull-up-only sharpen H-fbabeb50 (1.50 on 76), and the lift is smaller. Second, held-up chase pull-ups dropped back to hurdles scored 1.38 on 150 (p 0.17). So fences being taken away also shows a lift. Read together, the thread points away from 'a held-up pull-up hides ability' and towards 'a bad last run in the other code is over-read'. For chasing debutants this sits against the killed H124: fit, winning hurdlers on their chase debut scored 0.924. The market seems to price good hurdle form properly and under-price bad hurdle form. Two cautions. Both new pools are thin and neither clears p 0.05. The chase-to-hurdle cut has all 12 winners under 16/1 (0 from 94 at 16/1 or bigger) and a BSP ROI of -18.9%, the same under-16/1 concentration seen across the whole thread. And the debutant effect is GB 1.43 on 187 against IRE 0.91 on 148. That split came from a scan, so it is a lead, not a finding, and it should not be used to slice the rule.
investigator v1
Answered17 Oct, 02:39
Within the 226 jumps-to-jumps switchers, is the lift in hurdle-to-chase (a first try over fences) or chase-to-hurdle (fences taken away), and does a jumping error in the pull-up matter?
Both directions carry a lift: hurdle-to-chase held-up pull-ups scored 1.50 on 76 and chase-to-hurdle 1.38 on 150, so the cause looks like the change of code rather than a first try over fences. The jumping-error part is still untested.
Is the hurdle-to-chase lift about the pull-up hiding a horse's ability, or does any heavily beaten last hurdle run before a first chase beat its price, pulled up or not?
The evidence leans towards the broader effect: chase debutants after any hurdle beating of 20 lengths or more, pull-ups excluded, scored 1.24 on 335 (control 0.97). The pull-up is not needed, though the pull-up cut ran higher (1.50 on 76) and neither is conclusive.
H-7beeabeb (Test-mechanism: held-up chase pull-ups dropped back to hurdles) survived falsification: 4 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 02:59from H-7beeabeb
Tested17 Oct, 02:43
perturbation: Perturb heavy-beating cut down to 16 lengths: A/E 1.09 on 405 bets, p = 0.31 (95%: 0.78 to 1.48), by period 1.29 · 0.91. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:43
control: Control: chase debutants beaten under 10 lengths over hurdles: A/E 0.91 on 229 bets, p = 0.75 (95%: 0.66 to 1.22), by period 0.94 · 0.89. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:43
control: Control: prominent or front-running chase pull-ups dropped to hurdles: A/E 0.85 on 76 bets, p = 0.69 (95%: 0.23 to 2.17), by period 0.66. fail
H-448cb451 (Rival: chase debutants after any heavy hurdle beating beat their price) passed discovery but was falsified: IRE A/E 0.906.
near miss0.50reasoned about 17 Oct, 02:59from H-448cb451
Tested17 Oct, 02:43
control: Control: debutants beaten 10 to under 20 lengths (dose-response): A/E 0.78 on 190 bets, p = 0.88 (95%: 0.46 to 1.23), by period 0.73 · 0.82. fail
H-2e457254 (Rival: the debutant lift is the yard's chase-debut skill, not misread ) pointed the predicted way but failed the bar: A/E 1.09 on 296 bets (p 0.352), control 0.99.
near miss0.61reasoned about 17 Oct, 02:59from H-2e457254
Observed17 Oct, 02:43
H-551aa0a1 (Test-mechanism: chase debutants beaten heavily last time over hurdles ): the country 'GB' carries the effect (A/E 1.86) while the rest score IRE 0.73.
concentration0.55reasoned about 17 Oct, 02:59from H-551aa0a1
Observed17 Oct, 02:43
H-2e457254 (Rival: the debutant lift is the yard's chase-debut skill, not misread ): the country 'GB' carries the effect (A/E 1.27) while the rest score IRE 0.79.
concentration0.55reasoned about 17 Oct, 02:59from H-2e457254
Observed17 Oct, 02:43
H-551aa0a1 (Test-mechanism: chase debutants beaten heavily last time over hurdles ) passed discovery as a niche: A/E 1.3 on 86 bets (p 0.201), same-race control 0.95.
result0.50reasoned about 17 Oct, 02:59from H-551aa0a1
Tested17 Oct, 02:43
broad rule: A/E 1.09 on 296 bets, p = 0.35 (95%: 0.73 to 1.56), by period 1.43 · 0.81. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:43
control: Two comparisons. (1) The lab's same-race control: the other runners in these chases should score close to 1.0 (: A/E 1.29 on 218 bets, p = 0.15 (95%: 0.80 to 1.97), by period 1.55 · 0.91. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 02:44
H-551aa0a1 (Test-mechanism: chase debutants beaten heavily last time over hurdles ) passed discovery but was falsified: IRE A/E 0.735; fewer than two periods with enough bets to compare (thin: the forward record will judge).
near miss0.50reasoned about 17 Oct, 02:59from H-551aa0a1
Tested17 Oct, 02:44
control: Mechanism control: heavily beaten debutants with no close run in last three: A/E 1.29 on 218 bets, p = 0.15 (95%: 0.80 to 1.97), by period 1.55 · 0.91. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:44
perturbation: Perturb thresholds looser: last beaten 16+, best of last three within 7.5: A/E 1.01 on 119 bets, p = 0.52 (95%: 0.56 to 1.66), by period 0.85. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 02:44
perturbation: Perturb thresholds tighter: last beaten 24+, best of last three within 5: A/E 1.08 on 62 bets, p = 0.46 (95%: 0.47 to 2.12), by period 0.84. pass
In all three chase-debutant cuts, GB scored 1.27 to 1.86 and IRE 0.73 to 0.91. Is the split explained by handicap status or novice status, rather than by the country?
The split now repeats across three overlapping cuts, so it is a pattern and not a single scan segment. If debutants in handicaps are above par in both countries, the mechanism is a lenient mark after bad hurdle form, and the country is a side effect of GB running more novice handicap chases. If Ireland stays below 1 within handicaps, the market itself differs by country. A descriptive split would read region, is_handicap, is_novice, last_beaten_lengths, chase_runs, last_race_type and sp_decimal bands. race_class is not usable for Ireland before 2026-07.
answeredpriority 0.55raised by O-5c620d47, O-4137a04c, O-2930585a, O-6a4e2da9
New question17 Oct, 02:59
Does the parent held-up pull-up rule and its code-switch cuts (H-070b2026, H-7beeabeb) show the same GB-over-IRE split, or is that split specific to chase debutants?
If the published H36 system also lives only in GB, every child cut inherits a country confound, and the forward record should be read by region. If the parent is even across countries, the debutant split is a separate phenomenon. A scan would read region, last_pulled_up, last_run_style, race_type and last_race_type.
openpriority 0.40raised by O-d31892ab, O-5c620d47, O-4137a04c
To answer: In all three chase-debutant cuts, GB scored 1.27 to 1.86 and IRE 0.73 to 0.91. Is the split explained by handicap status or novice status, rather than by the country?
because A horse beaten 20 lengths or more over hurdles has its rating dropped. When it first runs over fences in a handicap, it carries a mark set on form in a discipline it may not have suited. If the horse jumps fences better than hurdles, that mark is lenient. GB runs many novice handicap chases, so GB debutants would land in this setting more often than Irish ones.
and the market misses it because The form figures show heavy recent defeats, and punters read them as a lack of ability. The mark looks fair on hurdle evidence, and nothing in the form shows how the horse will jump fences.
predicting positive, A/E 1.15 to 1.30 on 200+ bets, with Ireland's handicap subset at or above 1.0
and it is wrong if A/E below 1.05, or IRE handicap debutants below 0.95 again. Either result would mean the country, not handicap status, carries the split.
race_type = 'Chase' AND COALESCE(chase_runs, 0) = 0 AND last_race_type = 'Hurdle' AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND is_handicap = 1
The held-up pull-up thread has split into two branches. The code-switch branch is holding. Held-up chase pull-ups that drop back to hurdles (H-7beeabeb) survived 4 of 5 attacks and are now frozen for their vault query. That sits alongside H-070b2026 (1.359 on 304), which is still in shadow. The chase-debutant branch is not holding, and it keeps failing in the same way. Every debutant cut is above par in GB and below par in Ireland:
- any heavy hurdle beating (H-448cb451): GB 1.43, IRE 0.91, killed on IRE;
- heavily beaten last time over hurdles (H-551aa0a1): 1.30 on 86, GB 1.86 on 40, IRE 0.73 on 46, killed on IRE and on thin periods;
- the yard-skill rival (H-2e457254): 1.09 on 296 (p 0.35), GB 1.27, IRE 0.79.
Three cuts with the same split is no longer just one scan's segment. It is a repeated pattern, but it has not been explained. The yard rival scored below the yard-free debutant niche (1.09 against 1.24 on 335). So the published chase-debut yard effect does not account for the debutant lift. Either the lift is a misread of hurdle form that only the GB market makes, or it is a property of the GB race mix, such as novice handicap chases where a mark set on bad hurdle form is lenient. The H-551aa0a1 BSP ROI of +59.5% rests on 2 winners at 16/1 or bigger and means nothing. Slicing to GB would be slicing a near miss. The right test is the race-mix rival named in advance: handicap status should carry the effect in both countries.
investigator v1
Answered17 Oct, 02:59
Is the chase-debutant lift after a heavy hurdle beating carried by yards that are good at chase debuts, so that it is really the published yard effect (H104, SYS-chase-debut-from-a-good-yard) and not a misread of form?
No. The yard-skill cut (H-2e457254) scored only 1.09 on 296 (p 0.35), below the yard-free debutant niche at 1.24 on 335. The good-yard effect does not carry the debutant lift.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:02
H-cc72aea8 (Rival: the debutant lift is handicap marks after heavy hurdle beatings) passed discovery as a niche: A/E 1.25 on 225 bets (p 0.155), same-race control 0.98.
result0.50reasoned about 17 Oct, 03:06from H-cc72aea8
Observed17 Oct, 03:02
H-cc72aea8 (Rival: the debutant lift is handicap marks after heavy hurdle beatings): the country 'GB' carries the effect (A/E 1.38) while the rest score IRE 0.37.
concentration0.55reasoned about 17 Oct, 03:06from H-cc72aea8
Tested17 Oct, 03:02
refinement: GB handicap debutants (checks that the GB lift survives once handicap status is fixed): A/E 1.38 on 182 bets, p = 0.073 (95%: 0.89 to 2.04), by period 1.44 · 1.30. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:02
refinement: Shorter than 16/1 (a lenient mark should show up at backable prices, not only through a few long-shot winners): A/E 1.22 on 144 bets, p = 0.19 (95%: 0.78 to 1.84), by period 1.52 · 0.90. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:02
control: Same-race control: the other runners in these handicap chases should score about 1.0 (0.95 to 1.02). If they al: A/E 1.22 on 110 bets, p = 0.30 (95%: 0.61 to 2.18), by period 1.87 · 0.94. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 03:06
Is the heavily beaten debutant lift a winter-season effect that fades in summer jumping? It scored 1.45 before late April 2026 and 1.02 after.
Every cut in this branch skews to the first period, and the under-16/1 subset fell to 0.90 late on. If summer jumping (firmer ground, weaker fields) carries no lift, the forward record will look worse in summer, and any system should be read by season. Columns: race_date month, going_group, field_size, region.
openpriority 0.30raised by O-40119b5d
New question17 Oct, 03:06
Among handicap chase debutants beaten heavily over hurdles, does the lift sit with horses whose official rating fell since the last run?
The lenient-mark mechanism makes a direct prediction about the mark, but no rule has read the rating change yet. If the lift is no higher where official_rating < last_official_rating than where the mark held, the mark story is dead and the lift is something else, for example how well the horse jumps fences. Columns: official_rating, last_official_rating (from 2025-11), chase_runs, last_beaten_lengths, is_handicap.
answeredpriority 0.45raised by O-40119b5d
New question17 Oct, 03:06
Do horses beaten 20 lengths or more last time over hurdles beat their price when running in a jumps handicap next, in both countries? Is the chase-debutant lift just a slice of that?
A perturbation of H-cc72aea8 that drops race_type='Chase' scored 1.13 on 3675 (p 0.022), GB 1.16 and IRE 1.06. Dropping the debut condition scored 1.16 on 531, IRE 1.09. If the broad effect is real, the narrow GB-only debutant cuts are noise around it, and the thread should move to the broader population. Columns: last_beaten_lengths, last_race_type, is_handicap, race_type, region, sp_decimal, last_is_handicap.
answeredpriority 0.60raised by O-40119b5d, O-b183d4a0
To answer: Do horses beaten 20 lengths or more last time over hurdles beat their price when running in a jumps handicap next, in both countries? Is the chase-debutant lift just a slice of that?
because A beating of 20 lengths or more often has a cause that has nothing to do with ability: wrong ground, wrong trip, a hard pace, or a horse already beaten and then eased. The handicapper then drops the mark. The horse runs next off a lower mark with whatever caused the beating gone.
and the market misses it because The form figures show a big distance beaten, and punters count it twice: once as lack of ability, and again by ignoring the mark drop that already allows for it. The excuse behind the beating rarely shows in the bare result.
predicting positive, A/E 1.06 to 1.13 on 2500+ bets, both GB and IRE at or above 1.03
and it is wrong if A/E below 1.03, or IRE below 1.0, or the lift only at 16/1 or bigger (A/E under 1.0 for the bands up to 14/1 combined)
race_type IN ('Hurdle', 'Chase') AND is_handicap = 1 AND last_race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20
To answer: Among handicap chase debutants beaten heavily over hurdles, does the lift sit with horses whose official rating fell since the last run?
because If the lift comes from a lenient mark, it needs the mark to have actually fallen since the heavy hurdle beating. Debutants whose rating fell carry a mark set on form in a discipline that did not suit them. Those whose mark held do not get that benefit.
and the market misses it because Punters read the distance beaten in the form. The mark change and its meaning for a horse switching to fences are not obvious on the racecard.
predicting positive, A/E 1.25 to 1.50 on 100+ bets, clearly above debutants whose mark did not fall
and it is wrong if A/E below 1.10, or no higher than the debutants whose official_rating >= last_official_rating
race_type = 'Chase' AND COALESCE(chase_runs, 0) = 0 AND last_race_type = 'Hurdle' AND last_beaten_lengths >= 20 AND is_handicap = 1 AND official_rating IS NOT NULL AND last_official_rating IS NOT NULL AND official_rating < last_official_rating
The chase-debutant branch of H36 has moved a long way from its origin. The pull-up is not needed (H-448cb451: 1.24 on 335, pull-ups excluded). Yard skill does not explain it (H-2e457254: 1.09). H-cc72aea8 was the lenient-mark rival: confine the debutants to handicaps and the lift should appear in Ireland too. It passed only as a niche: 1.25 on 225 (p 0.155), same-race control 0.98. Its own decisive checks went against it. Irish handicap debutants scored 0.37 (1 winner from 43 bets, 2.7 expected), while GB scored 1.38 on 182 (p 0.07). The non-handicap mechanism control scored 1.22 on 110, level with the handicap group. That nearly meets the rival's own falsifier, so handicap status does not carry the lift.
Country and handicap status are almost the same variable here. Of 187 GB debutants, 182 ran in handicaps. Of 148 Irish ones, 105 ran in non-handicaps, and those scored 1.08. A country-versus-handicap split therefore cannot be separated inside this population.
The neighbours of the rule taught more than the rule did. Dropping the chase-debut condition gives 1.16 on 531, and the country gap closes (GB 1.18, IRE 1.09). Dropping the chase condition altogether, which leaves no-chase-run horses beaten 20 lengths or more over hurdles and now running in handicaps, gives 1.13 on 3675 (p 0.022), with GB 1.16 and IRE 1.06. So the narrow GB-only debutant figure looks like a noisy slice of a broader effect: horses beaten heavily last time, running in handicaps.
The period decay deserves caution: 1.45 in the first period against 1.02 after April 2026, and 0.90 late on under 16/1. BSP ROI of +25% also sits on 3 winners at 16/1 or bigger.
investigator v1
Answered17 Oct, 03:06
In all three chase-debutant cuts, GB scored 1.27 to 1.86 and IRE 0.73 to 0.91. Is the split explained by handicap status or novice status, rather than by the country?
Handicap status does not explain the split. Within handicaps GB scored 1.38 on 182 and IRE 0.37 on 43, and non-handicap debutants (mostly IRE) scored 1.22 on 110, level with the handicap group. Country and handicap status are nearly collinear in this population. Novice status is still untested (see Q-33ece0c3).
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:09
control: Mechanism control: same debutant profile in non-handicap chases: A/E 1.22 on 110 bets, p = 0.30 (95%: 0.61 to 2.18), by period 1.87 · 0.94. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:09
control: Ingredient control: handicap chase debutants NOT heavily beaten last time: A/E 0.88 on 164 bets, p = 0.79 (95%: 0.59 to 1.25), by period 0.91 · 0.85. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:09
H-cc72aea8 (Rival: the debutant lift is handicap marks after heavy hurdle beatings) passed discovery but was falsified: IRE A/E 0.367.
near miss0.50reasoned about 17 Oct, 03:13from H-cc72aea8
Tested17 Oct, 03:09
partition: Partition: non-novice handicaps only (strip novice-handicap race conditions): A/E 1.15 on 159 bets, p = 0.31 (95%: 0.68 to 1.82), by period 1.46 · 0.81. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:09
H-1a646ca5 (Generalise: jumps handicappers beaten 20 lengths or more last time bea) passed discovery: A/E 1.08 on 6743 bets (p 0.041), same-race control 0.99.
result0.60reasoned about 17 Oct, 03:13from H-1a646ca5
Tested17 Oct, 03:09
refinement: Mark dropped and 14/1 or shorter: A/E 1.03 on 2,614 bets, p = 0.28 (95%: 0.92 to 1.16), by period 1.03 · 1.04. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:09
control: Same-race control: the other runners in the same jumps handicaps should score about 1.0 (0.95 to 1.02). If they: A/E 0.95 on 8,444 bets, p = 0.97 (95%: 0.90 to 1.00), by period 0.95 · 0.95. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:10
H-1a646ca5 (Generalise: jumps handicappers beaten 20 lengths or more last time bea) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 03:13from H-1a646ca5
Tested17 Oct, 03:10
control: Control: heavy beating but mark not dropped: A/E 1.07 on 1,125 bets, p = 0.27 (95%: 0.86 to 1.32), by period 1.18 · 0.89. pass
H-cc8aceed (Test-mechanism: handicap chase debutants whose mark dropped after a he) passed discovery as a niche: A/E 1.33 on 164 bets (p 0.125), same-race control 0.97.
result0.50reasoned about 17 Oct, 03:13from H-cc8aceed
Observed17 Oct, 03:10
H-cc8aceed (Test-mechanism: handicap chase debutants whose mark dropped after a he): the country 'GB' carries the effect (A/E 1.5) while the rest score IRE 0.42.
concentration0.55reasoned about 17 Oct, 03:13from H-cc8aceed
Tested17 Oct, 03:10
forward vs bar {'min_bets': 300, 'min_ae': 1.03, 'max_p': 0.2} on 0 bets. info
control: The test is set by the mechanism itself: handicap chase debutants after a hurdle beating of 20 lengths or more : A/E 1.24 on 27 bets, p = 0.44 (95%: 0.26 to 3.62). info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 03:13
Are big beaten distances in large fields less informative, so that the lift concentrates where last_field_size was big?
In big jumps fields, horses far down the order are often eased once beaten, so their beaten lengths overstate how bad they are. This is the same 'looked after' logic as H36. A test would split last_field_size and last_eased.
answeredpriority 0.35raised by O-58c96d7b
New question17 Oct, 03:13
Is the broad heavy-beating lift carried by runners at 16/1 or bigger (1.174 on 3156), against the family pattern that longshots sit at or below par?
Nearly half the bets in O-58c96d7b are at 16/1 or bigger, and they score above par. That is the opposite of every code-switch cut in this thread (Q-4986ef47). A test would compare sp_decimal bands within the rule against the same-race control. If the edge sits only in long prices, it may be a jumps longshot-bias artefact in handicaps, not a reading of form.
testingpriority 0.45raised by O-58c96d7b, O-f18ec549
New question17 Oct, 03:13
Within jumps handicappers beaten 20+ lengths last time, does the lift sit with horses whose official rating fell since that run, compared with horses whose mark held or rose?
The broad rule (O-58c96d7b) is frozen. The mark-drop debutant cut (O-6a459966) scored 1.33, but only on 164 bets, and it was GB-only like its killed parent (O-b03feb4f), so it never tested the mark directly. A test would split the broad set on official_rating vs last_official_rating (both available from Nov 2025). If the lift is no higher where the mark fell, the lenient-mark story dies and a form-misread story is left.
answeredpriority 0.55raised by O-58c96d7b, O-6a459966, O-b03feb4f
To answer: Are big beaten distances in large fields less informative, so that the lift concentrates where last_field_size was big?
because A held-up horse that never got into the race is often eased down. The beaten distance overstates its effort, and the run took little out of it.
and the market misses it because The form book records the full beaten distance, and the market treats 25 lengths eased as 25 lengths of inability.
predicting positive, A/E 1.12 to 1.22 on 1500+ bets
and it is wrong if A/E at or below the broad rule's 1.08, or below 1.05 against the same-race control
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_run_style IN ('held_up', 'midfield')
To answer: Among handicap chase debutants beaten heavily over hurdles, does the lift sit with horses whose official rating fell since the last run?
because A 20-length beating usually brings a cut in the mark. If the beating was a one-off rather than a decline, the horse now races off a lenient rating.
and the market misses it because Punters read the big beaten distance as current form and stay away, but the mark has already priced in the run as if it were true.
predicting positive, A/E 1.12 to 1.20 on 1500+ bets, above the mark-held subset
and it is wrong if A/E below 1.05, or no higher than beaten-20+ runners whose mark did not fall
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND official_rating IS NOT NULL AND last_official_rating IS NOT NULL AND official_rating < last_official_rating
To answer: Are big beaten distances in large fields less informative, so that the lift concentrates where last_field_size was big?
because In fields of 14 or more, horses well down the order are routinely eased, and traffic costs ground. A big beaten distance there is weak evidence of inability.
and the market misses it because Beaten lengths are read the same whatever the field size, so big-field beatings get over-penalised.
predicting positive, A/E 1.12+ on 1500+ bets, clearly above the small-field subset
and it is wrong if A/E no higher than heavily beaten runners from fields of 10 or fewer, which would rule the rival out
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND last_field_size IS NOT NULL AND last_field_size >= 14
The thread began with held-up pull-ups (H36, 1.198 on 1329). It moved through code switchers and then hurdle-to-chase debutants, and now lands on a much broader population. Jumps handicappers beaten 20 lengths or more last time scored 1.08 on 6743 bets (p 0.041) against a control of 0.99. Both countries are above par (GB 1.087, IRE 1.067), and the rule survived 6 of 6 attacks.
The narrow debutant cuts keep failing in Ireland: the handicap-debutant rival died at IRE 0.367, and the mark-drop niche scored 1.33 on 164 bets but IRE 0.42 on 39. So the chase-debut story looks like a noisy, GB-skewed slice of a general effect: the market over-penalises a big beaten distance.
The edge is thin (BSP ROI 7.8%), and its price shape differs from the code-switch cuts. Here 16/1 or bigger is above par (1.174 on 3156), and evens or shorter is also strong.
What the chain has taught: the 'nothing taken out of it' and 'misread bad run' mechanisms seem to generalise from pull-ups to heavy beatings. But we still do not know why. It could be the lenient mark, easing once beaten, or how big fields spread beaten distances. Each of those predicts a different split, and all three are tested below. The vault query on H-1a646ca5 is now the key check. The code-switch shadow cuts still have 0 forward bets.
investigator v1
Answered17 Oct, 03:13
Do horses beaten 20 lengths or more last time over hurdles beat their price when running in a jumps handicap next, in both countries? Is the chase-debutant lift just a slice of that?
Yes. Jumps handicappers beaten 20 lengths or more last time scored 1.08 on 6743 (p 0.041) with GB 1.087 and IRE 1.067, and survived 6 of 6 attacks. The GB-only chase-debutant lift (killed at IRE 0.367) looks like a noisy slice of this broader effect.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:18
H-cc8aceed (Test-mechanism: handicap chase debutants whose mark dropped after a he) passed discovery but was falsified: IRE A/E 0.425.
near miss0.50reasoned about 17 Oct, 03:26from H-cc8aceed
Tested17 Oct, 03:18
control: Control: same dropped mark after a heavy hurdle beating, but staying over hurdles: A/E 1.10 on 2,672 bets, p = 0.11 (95%: 0.94 to 1.27), by period 1.08 · 1.12. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:18
control: Control: chase-experienced horses after a heavy hurdle beating with a dropped mark: A/E 0.87 on 162 bets, p = 0.72 (95%: 0.45 to 1.52), by period 0.90 · 0.83. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:18
H-b262a1d9 (Test-mechanism: heavily beaten jumps handicappers whose mark has dropp) passed discovery as a niche: A/E 1.07 on 4709 bets (p 0.106), same-race control 1.0.
result0.50reasoned about 17 Oct, 03:26from H-b262a1d9
Tested17 Oct, 03:18
refinement: One-off beating: within 10 lengths in one of its last three runs: A/E 1.14 on 1,447 bets, p = 0.068 (95%: 0.96 to 1.33), by period 1.07 · 1.21. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:18
control: Same-race other runners should be near par, around 0.99, as in the frozen heavy-beating finding. The comparison: A/E 1.07 on 1,241 bets, p = 0.25 (95%: 0.87 to 1.31), by period 1.20 · 0.89. info
H-23b2b028 (Sharpen: heavily beaten held-up handicappers, nothing taken out of the) pointed the predicted way but failed the bar: A/E 1.06 on 4329 bets (p 0.14), control 1.0.
near miss0.57reasoned about 17 Oct, 03:26from H-23b2b028
Tested17 Oct, 03:18
broad rule: A/E 1.06 on 4,329 bets, p = 0.14 (95%: 0.95 to 1.19), by period 1.07 · 1.06. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:18
refinement: Big-field beating came in a handicap (traffic-heavy races): A/E 1.17 on 1,691 bets, p = 0.044 (95%: 0.98 to 1.40), by period 1.18 · 1.16. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:18
H-d97a030e (Rival: big-field beatings, not the heavy beating itself, carry the lif) passed discovery: A/E 1.19 on 2097 bets (p 0.022), same-race control 0.98.
result0.60reasoned about 17 Oct, 03:26from H-d97a030e
Tested17 Oct, 03:18
control: The rival claim is that field size carries the lift, so the key comparison is jumps handicappers beaten 20+ len: A/E 1.02 on 3,281 bets, p = 0.37 (95%: 0.90 to 1.15), by period 1.02 · 1.03. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:21
H-b262a1d9 (Test-mechanism: heavily beaten jumps handicappers whose mark has dropp) survived falsification: 5 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 03:26from H-b262a1d9
Tested17 Oct, 03:21
perturbation: Perturb beaten threshold to 24 lengths: A/E 1.12 on 3,842 bets, p = 0.032 (95%: 0.99 to 1.25), by period 1.08 · 1.18. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:21
partition: One-off beating: within 10 lengths in its last three runs: A/E 1.14 on 1,447 bets, p = 0.068 (95%: 0.96 to 1.33), by period 1.07 · 1.21. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:21
H-d97a030e (Rival: big-field beatings, not the heavy beating itself, carry the lif) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 03:26from H-d97a030e
Tested17 Oct, 03:21
partition: Partition: handicap hurdles only: A/E 1.19 on 1,559 bets, p = 0.041 (95%: 0.98 to 1.44), by period 1.18 · 1.21. pass
Is the big-field heavy-beating lift a step down in competition (a smaller, weaker race today) rather than beaten lengths overstated by the size of the field?
The big-field rule (O-93b1b6e2) is IRE-heavy and lives under 16/1, which suggests festival-type handicaps. If the lift sits where today's field_size is much smaller than last_field_size, or where race_class is a drop from last_race_class (GB only; IRE race_class only from 2026-07), it is a competition effect. If it is flat across today's field size, the misleading-margin story stands. Columns: field_size, last_field_size, last_race_class, race_class, region.
answeredpriority 0.55raised by O-93b1b6e2, O-09d8e026
New question17 Oct, 03:26
Does the heavy-beating family hold two separate lifts: big last fields under 16/1, and small last fields at 16/1 or bigger?
Big last fields scored 1.274 up to 14/1 but 0.934 at 16/1 or bigger. Small last fields were near par up to 14/1 (1.045) but 1.242 at 16/1 or bigger. The cut-mark group is also 1.252 at 16/1 or bigger. A descriptive cross of sp_decimal bands by last_field_size bands within the frozen 6743-bet rule would show whether the longshot lift is a separate, general longshot effect. Columns: sp_decimal, last_field_size, last_beaten_lengths.
openpriority 0.45raised by O-93b1b6e2, O-2b2e7356, O-627a99c1
New question17 Oct, 03:26
Does a heavy beating in a big field mislead the market outside handicaps too, for example novice and maiden hurdlers?
The easing and traffic mechanism concerns the last race's field, not today's race conditions, so it should not need is_handicap = 1 today. Columns: is_handicap, is_novice, is_maiden, last_field_size, last_beaten_lengths, race_type.
To answer: Is the big-field heavy-beating lift a step down in competition (a smaller, weaker race today) rather than beaten lengths overstated by the size of the field?
because Horses beaten in big, competitive festival-type handicaps often race next in ordinary, smaller handicaps. There they meet weaker rivals, so the lift could come from the easier race and not from the margin misleading anyone.
and the market misses it because If the rival is right, the market under-weights the drop in race strength because the horse's last figure looks so bad. If it is wrong, the lift should be the same whatever today's field size.
predicting positive, A/E 1.25 or more on 500+ bets if the rival holds, clearly above big-field beaten runners now in fields of 14 or more
and it is wrong if A/E at or below the big-field parent's 1.187, or no higher than the same group running in fields of 14 or more today. That would rule the rival out and keep the misleading-margin mechanism.
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size <= 10
To answer: Is the broad heavy-beating lift carried by runners at 16/1 or bigger (1.174 on 3156), against the family pattern that longshots sit at or below par?
because In big, strung-out jumps handicaps the margins pile up all the way down the field. A horse finishing in the top half can still be beaten 20 lengths without having run badly, so its margin overstates how much it was inferior.
and the market misses it because Form readers and price compilers lean on beaten lengths. A top-half finish beaten 20 lengths reads as a bad run when it was a fair one in a strung-out race.
predicting positive, A/E 1.20 to 1.35 on 400+ bets, at or above the big-field parent's 1.187
and it is wrong if A/E below 1.10, or no higher than big-field heavily beaten runners that finished in the bottom half (last_position * 2 > last_field_size)
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_position IS NOT NULL AND last_position * 2 <= last_field_size
To answer: Does a heavy beating in a big field mislead the market outside handicaps too, for example novice and maiden hurdlers?
because Easing and traffic in a 14+ runner field inflate the margin whatever today's race conditions, so the effect should not depend on the next race being a handicap.
and the market misses it because Novice and maiden hurdlers are priced heavily off their visible form figures and margins, with no handicap mark to anchor them, so an inflated margin should hurt their price at least as much.
predicting positive, A/E 1.10 to 1.25 on 600+ bets, in both GB and IRE
and it is wrong if A/E below 1.05, or IRE below 0.95, or no higher than non-handicap runners beaten 20+ lengths in last fields of 10 or fewer
is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14
The thread began with held-up pull-ups. It widened to jumps handicappers beaten 20+ lengths last time (1.08 on 6743, frozen), and this round tested three explanations for that lift. Field size carries most of it, and the lift rises with field size: last fields of 14 or more scored 1.187 on 2097 (GB 1.215, IRE 1.171, both periods 1.21/1.16, 6/6 attacks held). Fields of 11 to 13 scored 1.116 on 1509 and fields of 10 or fewer scored 1.022 on 3281, with same-race controls near 0.98-1.0. The big-field lift sits under 16/1 (1.274 up to 14/1, 1.341 up to 7/1, against 0.934 at 16/1 or bigger). The small-field group's only above-par band is 16/1 or bigger (1.242). So the parent's longshot lift and its competitive-price lift look like two different things. The mark-cut mechanism failed its own test even though the rule froze (5/6). Cut marks scored 1.069 on 4709 and held or raised marks 1.075 on 1241. Under 16/1 the cut group was only 1.038. The mark-drop debutant cut was killed at IRE 0.425. The held-up 'nothing taken out' story adds nothing at this level either: 1.065 on 4329 is below the parent's 1.08. What we now believe: large fields make beaten lengths a misleading guide, and the market reads them as if they were not. Neither the handicap mark nor the running style is the cause. The open points are whether the big-field lift is really a step down in competition, and whether it reaches beyond handicaps. Note that 1591 of the 2097 big-field bets are Irish. 18+ fields gave a weak 1.029 on 649, so the dose-response does not keep climbing past 14 runners.
investigator v1
Answered17 Oct, 03:26
Among handicap chase debutants beaten heavily over hurdles, does the lift sit with horses whose official rating fell since the last run?
No. The mark-drop debutant cut was killed at IRE A/E 0.425 (GB 1.50, a single-country effect). The broader mark test found cut and uncut marks level (1.069 vs 1.075), so the rating change is not the mechanism.
Are big beaten distances in large fields less informative, so that the lift concentrates where last_field_size was big?
Largely yes. Heavy beatings in last fields of 14 or more scored 1.187 on 2097, fields of 11 to 13 scored 1.116 on 1509, and fields of 10 or fewer scored 1.022 on 3281. The lift rises with field size and held in both countries and both periods.
Within jumps handicappers beaten 20+ lengths last time, does the lift sit with horses whose official rating fell since that run, compared with horses whose mark held or rose?
No. Heavily beaten jumps handicappers with a cut mark scored 1.069 on 4709, against 1.075 on 1241 for a held or raised mark. Under 16/1 the cut group was only 1.038, so the mark cut does not carry the lift.
Tested17 Oct, 03:29
broad rule (niche): A/E 1.15 on 526 bets, p = 0.15 (95%: 0.88 to 1.48), by period 1.19 · 1.07. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:29
H-91a89a7c (Rival: the big-field lift is a drop into a smaller race today, not a m): the price band 'up to 7/1' carries the effect (A/E 1.51) while the rest score up to 3/1 0.82, up to 14/1 0.98, 16/1 or bigger 0.54.
concentration0.55reasoned about 17 Oct, 03:31from H-91a89a7c
Observed17 Oct, 03:29
H-91a89a7c (Rival: the big-field lift is a drop into a smaller race today, not a m) passed discovery as a niche: A/E 1.15 on 526 bets (p 0.154), same-race control 0.99.
result0.50reasoned about 17 Oct, 03:31from H-91a89a7c
Tested17 Oct, 03:29
refinement: Under 16/1 today (the band where the big-field lift sits): A/E 1.20 on 389 bets, p = 0.10 (95%: 0.91 to 1.55), by period 1.23 · 1.13. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:29
refinement: Handicap to handicap only (like-for-like drop in race strength): A/E 1.12 on 434 bets, p = 0.23 (95%: 0.83 to 1.48), by period 1.10 · 1.15. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:29
control: The deciding comparison is the same heavily beaten big-field group running today in fields of 14 or more. If th: A/E 1.24 on 1,136 bets, p = 0.054 (95%: 0.95 to 1.60), by period 1.07 · 1.38. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:30
control: Control without the big last field: heavy beating in a small race, now in a small race: A/E 1.07 on 1,870 bets, p = 0.20 (95%: 0.92 to 1.23), by period 1.06 · 1.09. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:30
control: Dose-response: mid-size field today (11-13) should sit between rule and big-field control: A/E 1.15 on 435 bets, p = 0.24 (95%: 0.78 to 1.64), by period 1.44 · 0.75. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:30
perturbation: Perturbation tightened: beaten 24+, last field 16+, today 8 or fewer: A/E 1.29 on 116 bets, p = 0.19 (95%: 0.73 to 2.09), by period 1.07. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:30
H-91a89a7c (Rival: the big-field lift is a drop into a smaller race today, not a m) passed discovery but was falsified: driven by one price band (up to 7/1): the rest score 0.88.
near miss0.50reasoned about 17 Oct, 03:31from H-91a89a7c
Tested17 Oct, 03:30
perturbation: Perturbation loosened: beaten 16+, last field 12+, today 11 or fewer: A/E 1.08 on 1,393 bets, p = 0.19 (95%: 0.91 to 1.27), by period 1.13 · 0.97. pass
H-9ccf4a8d (Test-mechanism: big-field beatings where the horse still finished in t) passed discovery: A/E 1.29 on 642 bets (p 0.03), same-race control 0.98.
result0.60reasoned about 17 Oct, 03:31from H-9ccf4a8d
Tested17 Oct, 03:30
refinement: Top third finish (stronger form of the mechanism): A/E 1.63 on 161 bets, p = 0.018 (95%: 1.03 to 2.44), by period 1.83 · 1.35. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:30
control: Same-race other runners should be near par (about 0.99, as in the parent's control), showing the lift is not a : A/E 1.12 on 1,455 bets, p = 0.15 (95%: 0.90 to 1.39), by period 1.11 · 1.14. info
H-f316faf4 (Generalise: big-field heavy beatings mislead in non-handicap jumps rac): the country 'GB' carries the effect (A/E 1.41) while the rest score IRE 0.93.
concentration0.55reasoned about 17 Oct, 03:31from H-f316faf4
Observed17 Oct, 03:30
H-f316faf4 (Generalise: big-field heavy beatings mislead in non-handicap jumps rac) pointed the predicted way but failed the bar: A/E 1.06 on 1684 bets (p 0.322), control 0.97.
near miss0.55reasoned about 17 Oct, 03:31from H-f316faf4
Tested17 Oct, 03:30
neighbour: >= 20 -> 16 (-20%): A/E 1.09 on 5,750 bets, p = 0.031 (95%: 1.00 to 1.19), by period 1.04 · 1.17. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:30
control: Control: top-half finish beaten 20+ lengths in small fields (8-13): A/E 0.92 on 614 bets, p = 0.76 (95%: 0.69 to 1.19), by period 0.82 · 1.14. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:30
H-9ccf4a8d (Test-mechanism: big-field beatings where the horse still finished in t) passed discovery but was falsified: 2 of 5 attacks broke it.
near miss0.50reasoned about 17 Oct, 03:31from H-9ccf4a8d
Tested17 Oct, 03:30
control: Control: big-field top-half finishers beaten under 20 lengths: A/E 0.98 on 1,437 bets, p = 0.60 (95%: 0.83 to 1.16), by period 0.96 · 1.01. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:30
partition: Partition: hurdles only (strung out without jumping errors): A/E 1.25 on 397 bets, p = 0.11 (95%: 0.88 to 1.71), by period 1.21 · 1.31. pass
Does the market's over-penalising of big-field heavy beatings also appear in Flat handicaps, where big fields are routine?
If the cause is that beaten lengths balloon in big fields once riders stop pushing, it should not depend on the code. The test would read race_type, last_race_type, last_field_size, last_beaten_lengths, is_handicap and sp_decimal.
testingpriority 0.40raised by O-6a8bcbc8, O-52f2a04d
New question17 Oct, 03:31
Is the big-field lift really a drop from a higher-class or festival handicap, rather than inflated beaten lengths?
The field-size-drop rival failed. A class or quality drop is a different rival: big fields cluster at valuable meetings. The test would read last_race_class against race_class (GB only; IRE race_class only from 2026-07), plus last_official_rating and official_rating, within the big-field rule.
testingpriority 0.40raised by O-4f7f9b7a, O-ac67c459, O-6a8bcbc8
New question17 Oct, 03:31
Is the big-field heavy-beating lift specific to handicaps because big-field handicap beatings overstate the gap, while big-field non-handicaps are mostly no-hopers in maidens and novices?
The non-handicap generalisation failed (1.057 on 1684, IRE 0.93, 74% of bets at 16/1 or bigger), but GB scored 1.41 on 291. A descriptive scan of is_novice, is_maiden, last_is_handicap and sp_decimal bands within last_field_size >= 14 and last_beaten_lengths >= 20 would show whether the failure is just a price-mix effect. Read it as a lead only.
To answer: Is the big-field lift really a drop from a higher-class or festival handicap, rather than inflated beaten lengths?
because Big fields cluster in valuable, higher-class handicaps. A horse well beaten there may simply face weaker rivals today.
and the market misses it because If the class drop explains it, the market misses the drop rather than misreading the lengths. Class drops are visible, though, so this rival is expected to score no higher than the parent.
predicting positive, A/E 1.20 to 1.35 on 300+ bets if class carries the lift
and it is wrong if A/E at or below the parent's 1.187, which would rule out class as the carrier
region = 'GB' AND is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND last_field_size IS NOT NULL AND last_field_size >= 14 AND last_race_class IS NOT NULL AND race_class IS NOT NULL AND last_race_class < race_class
To answer: Does the market's over-penalising of big-field heavy beatings also appear in Flat handicaps, where big fields are routine?
because In big fields, horses whose chance has gone are not ridden out, and traffic compounds the losses. The beaten distance therefore overstates the true gap, more so as the field gets bigger.
and the market misses it because The form book records the lengths. Punters read a 10-length beating in a 20-runner handicap as poor form, but the same performance in an 8-runner race would have been a few lengths.
predicting positive, A/E 1.06 to 1.12 on 1500+ bets, with the lift under 16/1
and it is wrong if A/E at or below 1.02 on 1000+ bets, or the lift only at 16/1 or bigger
race_type = 'Flat' AND is_handicap = 1 AND last_race_type = 'Flat' AND last_field_size >= 14 AND last_beaten_lengths >= 8 AND sp_decimal <= 15
The thread began with held-up pull-ups. It moved to heavy beatings (20 lengths or more) in jumps handicaps: 1.081 on 6743. Its one sharpening that worked was a big last field (14 or more runners): 1.187 on 2097, which survived 6 of 6 attacks and is frozen for the vault. This round tested three ideas around that big-field result and none of them improved on it. (1) Within the big-field cut, horses that still finished in the top part of the field scored 1.29 on 642 (p 0.03). That is higher than the parent, but 2 of 5 attacks broke it, so it is not a robust sharpening. (2) Opening the rule to non-handicaps failed: 1.057 on 1684 (p 0.32), IRE 0.93 on 1393, BSP ROI -46%, and 1254 of the bets were at 16/1 or bigger. So the effect is a handicap phenomenon, not a general one about big-field beatings. GB non-handicaps at 1.41 on 291 is one segment among several, not a finding. (3) The rival 'it is just a drop into a smaller race today' scored 1.15 on 526 (p 0.15). That is lower than the parent, and it was killed because only the up-to-7/1 band carried it (the rest 0.88). The drop in field size does not explain the lift. The pattern that repeats across all the big-field cuts is that 16/1 or bigger sits below par: 0.854, 0.826 and 0.544. The parent's 1.274 to 1.341 under 14/1 is where the effect lives. What we believe now: in handicaps, a heavy beating in a big field is over-penalised by the market for horses priced under about 14/1. Nothing has yet shown why. The frozen parent's vault and forward record are the real test.
investigator v1
Answered17 Oct, 03:31
Is the big-field heavy-beating lift a step down in competition (a smaller, weaker race today) rather than beaten lengths overstated by the size of the field?
No. The smaller-field-today cut scored 1.15 on 526, below the parent's 1.187, and was killed for resting on the up-to-7/1 band alone (rest 0.88). A step down in field size does not carry the big-field lift.
H-7358e6fb (Replicate: Flat handicappers heavily beaten in a big field last time b) was predicted positive but came out the other way: A/E 0.95 on 1560 bets (p 0.29 for the opposite direction).
surprise0.80reasoned about 17 Oct, 03:37from H-7358e6fb
Tested17 Oct, 03:35
broad rule (niche): A/E 1.25 on 176 bets, p = 0.16 (95%: 0.81 to 1.84), by period 1.30. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Killed17 Oct, 03:35
Replicate: Flat handicappers heavily beaten in a big field last time beat their priceDiscovery → KILLEDfails: A/E 0.954, needs 1.05; p 0.7359, needs 0.05 or lower; periods [1.11, 0.91], each needs 1.0; A/E 0.954 against same-race control 1.013, needs 0.03 above (not a niche pass: A/E 0.954 below 1.05; p 0.7359065308960852 above 0.5; does not beat the same-race control 1.013; a period points the other way [1.11, 0.91])
Observed17 Oct, 03:35
H-0bdb9150 (Rival: the big-field lift is a class drop out of valuable handicaps) passed discovery as a niche: A/E 1.25 on 176 bets (p 0.159), same-race control 0.98.
result0.50reasoned about 17 Oct, 03:37from H-0bdb9150
Tested17 Oct, 03:35
refinement: dropping out of a valuable (class 1-3) race: A/E 1.15 on 99 bets, p = 0.33 (95%: 0.63 to 1.93), by period 1.17. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:35
control: The key comparison is the parent's GB big-field heavy beatings that do NOT drop in class (same or higher class : A/E 1.27 on 270 bets, p = 0.13 (95%: 0.84 to 1.85), by period 1.32 · 1.11. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 03:37
Does a class drop carry the heavy-beating lift independently of last-time field size in GB jumps handicaps?
The class-drop cut scored 1.25 on 176, but the CI is 0.81 to 1.84 and it overlaps the parent. Widening it to every field size gives enough bets to separate class drop from field size. A test would read race_class against last_race_class (GB only), plus last_field_size and last_beaten_lengths.
testingpriority 0.40raised by O-3d6939aa
New question17 Oct, 03:37
Is the big-field heavy-beating lift jumps-only because jumps margins are inflated by easing and tiredness once beaten, which Flat races largely lack?
The Flat replicate scored 0.954 on 1560 against 1.187 for jumps, so the mechanism must be specific to jumps. A test would compare last_eased, last_never_involved and last_distance_yards within the jumps big-field rule (last_beaten_lengths >= 20, last_field_size >= 14) for jumps handicaps.
To answer: Is the big-field heavy-beating lift jumps-only because jumps margins are inflated by easing and tiredness once beaten, which Flat races largely lack?
because In big-field jumps handicaps, a beaten horse is often eased right down over the last half-mile. The recorded margin then piles up lengths that say nothing about its ability. Flat jockeys ease less, and Flat trips are shorter, which fits the Flat replicate's failure.
and the market misses it because The form figures and the beaten distance show a 25-length hammering. The fact that it was eased sits in the comment, and the price follows the margin.
predicting positive, A/E 1.20 to 1.35 on 400+ bets, above the big-field parent's 1.187
and it is wrong if A/E at or below 1.187, or no higher than the same rule with last_eased = 0
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_eased = 1
To answer: Does a class drop carry the heavy-beating lift independently of last-time field size in GB jumps handicaps?
because A heavy beating in a better race reflects the opposition, not the horse's ability. Today it meets weaker rivals off a mark that is often unchanged.
and the market misses it because The punter sees a 20-length defeat. The class of the race that produced it is less visible than the margin.
predicting positive, A/E 1.10 to 1.20 on 800+ bets, above the broad parent's 1.081
and it is wrong if A/E at or below 1.081, which would mean the class drop adds nothing and field size is the driver
region = 'GB' AND is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND race_class IS NOT NULL AND last_race_class IS NOT NULL AND race_class > last_race_class
The heavy-beating chain now has a clear outline. The broad jumps handicap rule (beaten 20+ lengths last time) scored 1.081 on 6743. Its big-field cut (last field 14+) is stronger, 1.187 on 2097, and that lift sits under 16/1. Two new tests narrow it further. First, the Flat replicate came out the opposite way to its prediction: 0.954 on 1560, GB 0.939, IRE 0.978, BSP ROI -9%. Whatever makes big-field heavy beatings mislead is a jumps phenomenon, not a general feature of beaten lengths in big fields. By rule 10, that points to something particular to jumps races, such as tired horses being eased over long trips once beaten, or jumping errors, which pile up lengths that say little about ability. The Flat has much less of this. Second, the class-drop rival passed only as a GB niche: 1.25 on 176 (CI 0.81 to 1.84, p 0.16), against a same-race control of 0.98. It is not distinguishable from the parent's 1.187, so a class drop neither explains the lift nor is ruled out. The sample is too thin to separate them. Rivals now ruled out: a smaller field today, the held-up style and the mark drop. Still standing: inflated jumps margins. None of the SHADOW rules has any forward bets yet, so everything here is still discovery-period evidence.
investigator v1
Tested17 Oct, 03:42
perturbation: Loosen thresholds: beaten 16+ lengths, last field 12+: A/E 1.27 on 395 bets, p = 0.049 (95%: 0.96 to 1.65), by period 1.18 · 1.58. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:42
H-0bdb9150 (Rival: the big-field lift is a class drop out of valuable handicaps) survived falsification: 4 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 03:44from H-0bdb9150
Tested17 Oct, 03:42
partition: Not from a premier race: last race class 2 or lower (excludes class 1 festival handicaps): A/E 1.20 on 139 bets, p = 0.26 (95%: 0.71 to 1.89), by period 1.28. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:42
control: Control: heavy beating plus class drop, but last field under 14: A/E 1.14 on 1,095 bets, p = 0.098 (95%: 0.94 to 1.37), by period 1.10 · 1.23. pass
H-52702d58 (Test-mechanism: big-field jumps beatings where the horse was eased car) pointed the predicted way but failed the bar: A/E 1.76 on 18 bets (p 0.314), control 0.96.
near miss0.60reasoned about 17 Oct, 03:44from H-52702d58
Tested17 Oct, 03:42
broad rule: A/E 1.76 on 18 bets, p = 0.31 (95%: 0.21 to 6.35). fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:42
H-6f7a705a (Generalise: GB jumps handicappers heavily beaten in a higher class, an) passed discovery as a niche: A/E 1.16 on 1271 bets (p 0.053), same-race control 0.98.
result0.50reasoned about 17 Oct, 03:44from H-6f7a705a
Tested17 Oct, 03:42
forward vs bar {'min_bets': 100, 'min_ae': 1.0, 'max_p': 0.2} on 0 bets. info
control: The other runners in the same races should come out near A/E 1.0. If they also beat their prices, the lift belo: A/E 1.07 on 2,834 bets, p = 0.15 (95%: 0.94 to 1.22), by period 1.07 · 1.07. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 03:44
Among GB jumps handicappers beaten 20+ lengths, does dropping in class beat staying at the same class, or is the 1.155 just the GB heavy-beating base rate?
H-6f7a705a scored 1.155 on 1271 but has no internal control for class. Compare race_class > last_race_class against race_class = last_race_class within region = 'GB', is_handicap = 1, last_beaten_lengths >= 20, reading A/E by sp_decimal band. If the two are equal, class relief is not the mechanism.
answeredpriority 0.55raised by O-bc2f6f2e, O-b650f2b7
New question17 Oct, 03:44
Is last_eased sparsely recorded in jumps results, so that easing tests return tiny samples? Would last_never_involved or last_one_paced be usable proxies for a horse not ridden out once beaten?
H-52702d58 found only 18 big-field jumps handicap bets with an eased flag, out of more than 2000 parent bets. That points to a coverage gap, not a rare event. A scan should count non-null and positive values of last_eased, last_never_involved and last_one_paced within last_beaten_lengths >= 20 AND last_field_size >= 14, by region and month.
answeredpriority 0.50raised by O-1d81d740
New question17 Oct, 03:44
Is the longshot lift in the class-drop group (1.32 on 461 at 16/1 or bigger) carried by runners from small last fields, matching the split asked about in Q-a18717ff?
Big-field cuts sit below par at 16/1 or bigger, but the class-drop group, at any field size, is above par there. A descriptive cross of last_field_size (<= 10 vs >= 14) by sp_decimal band within the H-6f7a705a rule would show whether the two lifts are really separate.
To answer: Among GB jumps handicappers beaten 20+ lengths, does dropping in class beat staying at the same class, or is the 1.155 just the GB heavy-beating base rate?
because If the lift comes from a heavy beating looking worse in the form book than the horse really is, it should not depend on whether the trainer drops the horse in class. If class relief is the cause, the same-class group should sit near par.
and the market misses it because Punters read a 20+ length beating as a horse out of form whatever the class. The margin, not the class move, is what puts them off. If the same-class group also beats its price, the market is mispricing the margin itself.
predicting positive, A/E 1.08 to 1.15 on 1500+ bets if the rival holds
and it is wrong if A/E at or below 1.03, or more than 0.08 below the class-drop group's 1.155, which would mean the class drop itself carries the lift
region = 'GB' AND is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND race_class IS NOT NULL AND last_race_class IS NOT NULL AND race_class = last_race_class
To answer: Is the big-field heavy-beating lift jumps-only because jumps margins are inflated by easing and tiredness once beaten, which Flat races largely lack?
because In big jumps fields, a horse that is never involved is usually allowed to come home in its own time once beaten. That inflates the beaten distance well beyond its real merit, which is the easing mechanism the 18-bet test was too narrow to measure.
and the market misses it because The form book shows a 20+ length beating and a 'never involved' comment, and both read as a horse with no ability. The margin reflects how little effort was asked for, not ability.
predicting positive, A/E 1.20 to 1.35 on 400+ bets, at or above the big-field parent's 1.187
and it is wrong if A/E below 1.10, or no higher than big-field heavily beaten runners with last_never_involved = 0
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_never_involved = 1
The H36 thread has become a story about heavy beatings in jumps handicaps. Two facts hold: the broad heavy-beating rule (1.081 on 6743) and its big-field cut (1.187 on 2097, 6/6 attacks survived). Every attempt to pin the big-field lift to one cause has stalled. A drop into a smaller field today was killed. A top-half finishing position was killed by attacks. The class-drop rival inside big fields (H-0bdb9150) is now frozen after 4 of 5 attacks, but it is too thin to say anything until its vault query. The direct test of the easing mechanism (H-52702d58) found only 18 qualifying bets (A/E 1.76, CI 0.21 to 6.35). So the rule was far too narrow, or last_eased is sparsely recorded. It tells us nothing about whether easing inflates the margins. The most informative new result is H-6f7a705a: GB jumps handicappers beaten 20+ lengths and now dropping in class, any field size, scored 1.155 on 1271 (p 0.053, control 0.98) and passed as a niche. Unlike every big-field cut, its 16/1-or-bigger band is above par (1.32 on 461), with 14/1-and-under also around 1.30. So the class-drop lift is not confined to the shorter prices the way the big-field lift is. That suggests two different sources: big-field margins that mislead (short prices) and class relief (all prices, longshots included). The open test is whether class drop adds anything over a plain GB heavy beating. That needs the same population with no class drop as a control.
investigator v1
Tested17 Oct, 03:50
perturbation: Tighter beating threshold: 25 lengths or more: A/E 1.18 on 1,013 bets, p = 0.057 (95%: 0.96 to 1.43), by period 1.19 · 1.15. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:50
partition: Small last-time field (<12): separate from the big-field lift: A/E 1.08 on 935 bets, p = 0.23 (95%: 0.87 to 1.33), by period 1.08 · 1.09. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:50
H-6f7a705a (Generalise: GB jumps handicappers heavily beaten in a higher class, an) passed discovery but was falsified: 2 of 5 attacks broke it; IRE: too few bets to judge.
near miss0.50reasoned about 17 Oct, 04:06from H-6f7a705a
Tested17 Oct, 03:50
control: Control: class droppers that were NOT heavily beaten (under 8 lengths): A/E 0.91 on 650 bets, p = 0.85 (95%: 0.76 to 1.09), by period 0.91 · 0.92. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Killed17 Oct, 03:50
Rival: the GB heavy-beating lift needs no class dropDiscovery → KILLEDfails: A/E 1.007, needs 1.05; p 0.4697, needs 0.05 or lower; A/E 1.007 against same-race control 1.008, needs 0.03 above (not a niche pass: A/E 1.007 below 1.05; does not beat the same-race control 1.008)
Tested17 Oct, 03:50
broad rule: A/E 1.32 on 1,215 bets, p = 0.0052 (95%: 1.07 to 1.62), by period 1.32 · 1.32. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:50
H-f956317c (Test-mechanism: big-field heavy beatings where the horse was never inv) passed discovery: A/E 1.32 on 1215 bets (p 0.005), same-race control 0.98.
result0.60reasoned about 17 Oct, 04:06from H-f956317c
Tested17 Oct, 03:50
control: Same-race other runners are expected near par (about 0.95 to 1.0), as in earlier big-field cuts (control 0.98).: A/E 1.00 on 863 bets, p = 0.51 (95%: 0.75 to 1.30), by period 1.06 · 0.92. info
perturbation: Loosen thresholds: beaten 16+ lengths, last field 12+: A/E 1.21 on 1,930 bets, p = 0.013 (95%: 1.02 to 1.43), by period 1.23 · 1.19. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 03:51
H-f956317c (Test-mechanism: big-field heavy beatings where the horse was never inv) survived falsification: 5 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:06from H-f956317c
Tested17 Oct, 03:51
perturbation: Tighten thresholds: beaten 24+ lengths, last field 16+: A/E 1.44 on 629 bets, p = 0.0095 (95%: 1.06 to 1.90), by period 1.41 · 1.47. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:51
control: Control: never involved but only moderately beaten (8 to 19 lengths): A/E 0.93 on 324 bets, p = 0.67 (95%: 0.60 to 1.37), by period 1.05 · 0.82. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 03:51
control: Control: heavy beating and never involved, but small last field (6 to 11): A/E 1.06 on 1,717 bets, p = 0.27 (95%: 0.88 to 1.26), by period 1.06 · 1.05. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 04:06
Why does the never-involved lift vanish when last time's field had 17+ runners (1.005 on 461), while 11+ gives 1.33 and 24+ lengths in 16+ gives 1.437?
A mechanism based on field size should grow with field size, not collapse at 17+. The 17+ fields may be festival or valuable handicaps, where a never-involved horse really was outclassed. A descriptive scan should read last_field_size bands, last_race_class, last_course_id and the month of the last run, before anyone proposes a rule.
openpriority 0.30raised by O-86a41341
New question17 Oct, 04:06
Is the never-involved lift strongest when the heavy beating was a one-off, meaning the horse finished close in one of its other recent runs?
The figures pay where a single run made the horse look worse than it is. If the margin was inflated because the horse was not asked for effort, horses with a recent close finish should be the most mispriced. A test reads best_last3_beaten, last_never_involved, last_beaten_lengths and last_field_size, and compares best_last3_beaten <= 6 against > 6.
testingpriority 0.40raised by O-86a41341, O-298ba57a
New question17 Oct, 04:06
Do Flat handicappers beaten 20+ lengths in a 14+ field and marked 'never involved' beat their price? The removed-race_type neighbour implies about 1.22 on roughly 539 non-jumps bets.
The jumps-only conclusion rests on H-7358e6fb, which had no comment flag and used 8+ lengths. Dropping race_type from H-f956317c gave 1.296 on 1754 bets against 1.324 on 1215 for jumps, which implies about 1.22 on the rest. If the Flat version holds, the mechanism is 'not ridden out once beaten', not jumps fatigue, and the lab's scope widens. A test reads race_type, last_race_type, is_handicap, last_beaten_lengths, last_field_size, last_never_involved and sp_decimal.
answeredpriority 0.50raised by O-86a41341, O-0cca5f69
To answer: Is the never-involved lift strongest when the heavy beating was a one-off, meaning the horse finished close in one of its other recent runs?
because If the margin was inflated because the horse was not ridden out, a horse that recently finished close has ability that the latest result hides. A horse with no close runs is more likely to be truly out of form, and for that horse 'never involved' is an accurate description.
and the market misses it because Prices are drawn mostly from the latest run, and a 20-length 'never involved' line looks worse than an older close finish looks good. The one-off bad run is overweighted.
predicting positive, A/E 1.35 to 1.60 on 300+ bets, above the parent's 1.324
and it is wrong if A/E no higher than the same rule with best_last3_beaten > 6, or A/E below 1.15
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_never_involved = 1 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6
To answer: Do Flat handicappers beaten 20+ lengths in a 14+ field and marked 'never involved' beat their price? The removed-race_type neighbour implies about 1.22 on roughly 539 non-jumps bets.
because In a big field, a horse that is never in contention is often allowed to coast home once its chance has gone, whatever the code. A 20+ length margin then measures how little was asked of it, not its ability. The earlier Flat test missed this because it had no comment flag and used 8+ lengths, which mixes in horses that were ridden out.
and the market misses it because The form line shows a heavy beating plus 'never involved', and readers count them as two pieces of bad evidence. They are one event, and the double penalty pushes the price out too far.
predicting positive, A/E 1.10 to 1.25 on 300+ bets
and it is wrong if A/E below 1.05, or no higher than the same rule with last_never_involved = 0, or the lift only at 16/1 or bigger
race_type = 'Flat' AND is_handicap = 1 AND last_race_type = 'Flat' AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_never_involved = 1
The thread has narrowed from a loose heavy-beating effect to a specific mechanism. Jumps handicappers beaten 20+ lengths scored 1.081 on 6743 bets. Restricting to fields of 14+ last time raised that to 1.187 on 2097. Every explanation for the big-field cut died until H-f956317c added the 'never involved' comment flag. Inside the same big-field set, flagged runners score 1.324 on 1215 bets (p 0.005, CI 1.07 to 1.62). The two periods are identical at 1.32 and 1.32, and GB (1.316) and IRE (1.329) agree. The unflagged remainder scores 1.001 on 863. So the whole big-field lift sits in horses that were never in the race and were allowed to come home in their own time.
Both ingredients are needed. A never-involved horse beaten only 8 to 19 lengths scores 0.927 on 324. A heavy beating with the flag in a field of 6 to 11 scores 1.058 on 1717. The rule survived 5 of 6 attacks and is frozen. Tightening the cut to 24+ lengths in fields of 16+ gives 1.437, and hurdles alone give 1.374. Two things weaken it. The 17+ field neighbour drops to 1.005 on 461, so field size does not work like a smooth dose. The GB part is thin: 338 bets, p 0.059, unstable across price bands.
The flag also removes the longshot drag that has run through this family. Flagged runners at 16/1 or bigger score 1.175, against 0.622 for unflagged ones. The GB class-drop generalisation (H-6f7a705a) died. Its 1.155 barely beat GB heavily beaten runners not dropping in class (1.071), and class droppers whose mark had not been cut scored 0.668 on 145. Class drop is not a separate carrier.
One lead cuts against an earlier conclusion. Removing race_type from the never-involved rule gives 1.296 on 1754 bets. The non-jumps remainder, roughly 539 bets that are mostly Flat handicaps, works out at about 1.22. The 'jumps only' lesson came from a Flat rule with no comment flag and only 8+ lengths, so it may be wrong.
investigator v1
Answered17 Oct, 04:06
Among GB jumps handicappers beaten 20+ lengths, does dropping in class beat staying at the same class, or is the 1.155 just the GB heavy-beating base rate?
Dropping in class adds little. The drop scored 1.155 on 1271 against 1.071 on 2834 for no drop, with overlapping CIs. Class droppers whose mark had not been cut scored 0.668, so the 1.155 is mostly the GB heavy-beating base rate plus mark relief.
Is last_eased sparsely recorded in jumps results, so that easing tests return tiny samples? Would last_never_involved or last_one_paced be usable proxies for a horse not ridden out once beaten?
Yes. In the big-field heavy-beating set, last_never_involved splits 1215 flagged against 863 unflagged, about the whole 2097-bet parent. It is well populated and usable where last_eased (18 bets) was not.
Tested17 Oct, 04:09
broad rule (niche): A/E 1.31 on 463 bets, p = 0.091 (95%: 0.88 to 1.88), by period 1.31 · 1.31. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:09
refinement: Below 16/1 (jumps lift sat below 16/1; tests that the edge is not a longshot artefact): A/E 1.06 on 160 bets, p = 0.43 (95%: 0.63 to 1.68), by period 1.24 · 0.99. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:09
H-c9d4fd57 (Replicate: Flat handicappers heavily beaten and never involved in a bi) passed discovery as a niche: A/E 1.31 on 463 bets (p 0.091), same-race control 0.98.
result0.50reasoned about 17 Oct, 04:11from H-c9d4fd57
Tested17 Oct, 04:09
broad rule: A/E 1.72 on 236 bets, p = 0.0026 (95%: 1.18 to 2.41), by period 1.71 · 1.73. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:09
control: Primary control is the same rule with last_never_involved = 0: Flat handicappers beaten 20+ lengths in a 14+ fi: A/E 1.02 on 232 bets, p = 0.51 (95%: 0.53 to 1.78), by period 0.91 · 1.04. info
H-baef90d8 (Sharpen: never-involved heavy beatings by horses with a recent close f) passed discovery: A/E 1.72 on 236 bets (p 0.003), same-race control 0.95.
result0.60reasoned about 17 Oct, 04:11from H-baef90d8
Tested17 Oct, 04:09
control: The other runners in the same races should score close to 1.0, which shows the lift is not a property of the ra: A/E 1.18 on 979 bets, p = 0.12 (95%: 0.90 to 1.51), by period 1.21 · 1.13. info
price_band: Cap the longshots at 33/1: is the lift just a few huge-price winners?: A/E 1.26 on 294 bets, p = 0.14 (95%: 0.83 to 1.85), by period 1.24 · 1.27. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:10
H-c9d4fd57 (Replicate: Flat handicappers heavily beaten and never involved in a bi) survived falsification: 4 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:11from H-c9d4fd57
Tested17 Oct, 04:10
perturbation: Tighten thresholds: 24+ lengths, 16+ runner field: A/E 1.49 on 165 bets, p = 0.12 (95%: 0.77 to 2.60), by period 1.76 · 1.34. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:10
control: Control: same flag and margin, but a small field last time (8-13): A/E 0.84 on 873 bets, p = 0.86 (95%: 0.57 to 1.18), by period 0.89 · 0.82. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:10
partition: Partition: a mile and a half or shorter (no staying handicaps): A/E 1.23 on 394 bets, p = 0.19 (95%: 0.78 to 1.84), by period 1.28 · 1.20. pass
perturbation: Tighter close-finish and margin cuts (best_last3 <= 4.5, beaten >= 24): A/E 1.83 on 177 bets, p = 0.0027 (95%: 1.21 to 2.66), by period 1.63 · 2.00. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:10
perturbation: Looser cuts (best_last3 <= 7.5, beaten >= 16, field >= 12): A/E 1.43 on 470 bets, p = 0.0070 (95%: 1.08 to 1.86), by period 1.65 · 1.26. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:10
partition: Close losers only: drop recent winners (best_last3_beaten > 0): A/E 2.00 on 146 bets, p = 0.0022 (95%: 1.26 to 3.03), by period 1.86 · 2.11. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:10
H-baef90d8 (Sharpen: never-involved heavy beatings by horses with a recent close f) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:11from H-baef90d8
Tested17 Oct, 04:10
control: Control: same profile without the never-involved flag: A/E 1.47 on 186 bets, p = 0.052 (95%: 0.92 to 2.23), by period 1.59 · 1.35. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 04:11
Is the jumps lift a rebound from one bad run after recent close form, holding without the never-involved flag and without a big last field?
In the H-baef90d8 attacks, close finishers without the flag scored 1.47 on 186, close finishers from 6 to 13 runner fields scored 1.334 on 517, and lowering the field cut to 11+ raised it to 1.77 on 410. If the close finish carries the lift, the flag and field-size conditions are only rough ways of finding the run, and the published rule should be rebuilt around best_last3_beaten. A test would read best_last3_beaten, last_beaten_lengths, last_never_involved and last_field_size within jumps handicaps.
answeredpriority 0.65raised by O-e041a181, O-e25c4c84
New question17 Oct, 04:11
Why is the close-finish heavy-beating lift so much stronger in Ireland (2.11 on 153) than in GB (1.30 on 83)?
The country split repeats across neighbours (IRE 2.36 at 24+ lengths against GB 1.14). Possible causes: Irish comment-writing style for 'never involved', more big-field handicaps in IRE, or Irish yards running horses down the weights. A descriptive scan would read region, last_field_size, last_never_involved, trainer and days_since_last_run.
openpriority 0.35raised by O-e041a181
New question17 Oct, 04:11
Does the one-bad-run rebound replicate on the Flat once recent close form is required, moving the Flat lift out of the 16/1+ band?
The Flat replicate's lift is entirely at 16/1 or bigger (11 wins). If the jumps mechanism is a one-off bad run after recent close form, requiring a close recent finish should pick out Flat runners with real ability, and they should be priced below 16/1. A test would read race_type, last_race_type, last_beaten_lengths, best_last3_beaten and sp_decimal.
testingpriority 0.40raised by O-276bf66a, O-463063c7, O-e041a181
To answer: Is the jumps lift a rebound from one bad run after recent close form, holding without the never-involved flag and without a big last field?
because A horse that finished within 6 lengths in one of its previous two runs has shown its current level. A single 20+ length beating after that is more often caused by something temporary (the trip, the ground, a bad experience, not being asked once beaten) than by a lasting loss of form. The never-involved flag and the big field may only have been rough ways of finding such runs.
and the market misses it because Prices and casual readers lean on the latest run. A 20-length beating is the most visible line in the form, and an older close finish is weighed less than it deserves, so the horse drifts.
predicting positive, A/E 1.30 to 1.50 on 700+ bets if the rival holds
and it is wrong if A/E at or below 1.177 (the no-close-finish flagged control). Then the close finish only works alongside the flag and the big field, and the rival is ruled out.
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6
To answer: Does the one-bad-run rebound replicate on the Flat once recent close form is required, moving the Flat lift out of the 16/1+ band?
because The same cause as over jumps: one bad run after close recent form usually has a temporary cause. Flat margins are compressed, so the cuts are scaled down: 12+ lengths counts as a heavy beating and within 3 lengths counts as close. These cuts were chosen before seeing any numbers.
and the market misses it because Flat punters are used to small margins, so a 12+ length defeat looks disastrous and dominates the price, even when the horse ran close in one of its two runs before that.
predicting positive, A/E 1.10 to 1.25 on 400+ bets, with the lift present below 16/1
and it is wrong if A/E below 1.05, or the sp_decimal < 17 subset at or below 1.0, which would mean the Flat effect is still only a longshot artefact
race_type = 'Flat' AND is_handicap = 1 AND last_race_type = 'Flat' AND last_beaten_lengths >= 12 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 3
The thread has moved from 'pulled up after being held up' to a narrower story: a heavy beating that doesn't reflect the horse's ability. The jumps chain so far: heavily beaten jumps handicappers scored 1.081 on 6743. In big last fields that rose to 1.187 on 2097. Adding the never-involved flag gave 1.324 on 1215, and unflagged runners scored 1.001. This round's close-finish sharpen (H-baef90d8) is the strongest cut yet: 1.72 on 236 bets, p 0.003, periods 1.71 and 1.73, same-race control 0.95. All 6 attacks held, and it improves on its own control (best_last3_beaten > 6: 1.177 on 979). Its attack results also shift the explanation. With the close finish in place, the flag adds less: unflagged close finishers scored 1.47 on 186 (p 0.052), just under the 1.5 kill line. The big field adds less too: last fields of 6 to 13 scored 1.334 on 517, and lowering the field cut to 11+ gave 1.77 on 410. The likeliest carrier is now 'one bad run after recent close form'. The big field and the flag are the earlier, cruder ways of finding that run. IRE (2.11) carries far more than GB (1.30), and that is still unexplained. The Flat replicate (H-c9d4fd57) passed only as a niche: 1.31 on 463, p 0.09. The flag matters on the Flat too (unflagged control 1.019), but the whole lift sits at 16/1 or bigger (2.13 on 303 bets from 11 wins). Below 16/1 it is 1.062 on 160. It falls to 0.932 when the last-field cut drops to 11+, and to 0.839 for small fields. So the Flat version is a fragile longshot niche, not a clean replicate. Both rules are frozen in discovery with 0 forward bets.
investigator v1
Answered17 Oct, 04:11
Do Flat handicappers beaten 20+ lengths in a 14+ field and marked 'never involved' beat their price? The removed-race_type neighbour implies about 1.22 on roughly 539 non-jumps bets.
Partly. Flat handicappers beaten 20+ lengths in a 14+ field and marked never involved score 1.31 on 463 (niche, p 0.09), against 1.019 unflagged. But 16/1 or bigger carries all of it (2.13 on 303, 11 wins), below 16/1 is 1.062 on 160, and the 11+ field neighbour is 0.932.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:12
H-332b49e9 (Rival: the lift is a one-bad-run rebound after close recent form, need) passed discovery: A/E 1.23 on 1600 bets (p 0.004), same-race control 0.98.
result0.60reasoned about 17 Oct, 04:13from H-332b49e9
Tested17 Oct, 04:12
refinement: no mark relief after the heavy beating: A/E 1.31 on 506 bets, p = 0.026 (95%: 1.00 to 1.68), by period 1.45 · 1.15. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:12
control: Same-race other runners should score about 1.0, showing the lift is not a property of the races these horses ru: A/E 1.02 on 5,287 bets, p = 0.34 (95%: 0.92 to 1.14), by period 1.05 · 0.98. info
Replicate: Flat handicappers with a close recent finish after one heavy beatingDiscovery → KILLEDfails: A/E 1.021, needs 1.05; p 0.4117, needs 0.05 or lower; periods [0.93, 1.04], each needs 1.0; A/E 1.021 against same-race control 1.0, needs 0.03 above (not a niche pass: A/E 1.021 below 1.05; does not beat the same-race control 1.0; a period points the other way [0.93, 1.04])
Tested17 Oct, 04:12
neighbour: >= 20 -> 16 (-20%): A/E 1.16 on 2,079 bets, p = 0.015 (95%: 1.01 to 1.32), by period 1.16 · 1.16. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 04:13
Does the close-finish rebound hold below 16/1, or is it carried by a few longshot winners?
16/1 or bigger scores 1.58 from 22 wins, and SP ROI (+2.4%) sits far below BSP ROI (+37.9%). Up to 14/1 scores 1.31 on 537 and up to 7/1 1.20 on 424. Bettability depends on the lift below 16/1. Columns: sp_decimal, sp_rank.
openpriority 0.40raised by O-10a4d6e5
New question17 Oct, 04:13
Among jumps handicappers beaten 20+ lengths, do recent winners (best_last3_beaten = 0) beat their price more than horses with a close placed finish, and why does this reverse the parent niche?
Excluding recent winners cut H-332b49e9 from 1.232 to 1.152 and GB to 1.031. That implies about 1.34 on 669 for winners, the reverse of the never-involved niche (2.00 without winners). A test would read best_last3_beaten, days_since_win, rating_change and last_race_class.
answeredpriority 0.55raised by O-10a4d6e5
New question17 Oct, 04:13
Is the one-bad-run rebound strongest when the heavy beating came on different ground or at a different trip from today, as a temporary-cause mechanism predicts?
The mechanism names trip and ground as temporary causes. If they are real, runners whose last run was on a different going_group or distance should score above those on the same conditions. Columns: last_going_group, going_group, last_distance_yards, distance_yards.
To answer: Among jumps handicappers beaten 20+ lengths, do recent winners (best_last3_beaten = 0) beat their price more than horses with a close placed finish, and why does this reverse the parent niche?
because A recent win is the clearest proof of current ability. The heavy beating that followed often came after a rise in the weights or class and a hard race, which is a temporary setback rather than a loss of form, so the horse recovers to near its winning level.
and the market misses it because Punters read a heavy defeat straight after a win as being found out by the handicapper, and let the horse drift, even though the winning form is only weeks old.
predicting positive, A/E 1.25 to 1.45 on 500+ bets, in both GB and IRE
and it is wrong if A/E at or below 1.15, or GB below 1.05, or no higher than the placed-but-not-won group (1.152)
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten = 0
To answer: Is the one-bad-run rebound strongest when the heavy beating came on different ground or at a different trip from today, as a temporary-cause mechanism predicts?
because If the 20-length beating had a temporary cause, the commonest causes are unsuitable ground or trip. A horse back on different going from the beating faces conditions closer to those of its close run, so the beating says least about today.
and the market misses it because The form line shows the margin prominently. The going of each run is shown but is rarely used to discount a beating, so the latest heavy defeat drags the price whatever the conditions.
predicting positive, A/E 1.25 to 1.45 on 500+ bets, above the same-going complement
and it is wrong if A/E at or below 1.15, or no higher than the same rule with last_going_group = going_group
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6 AND last_going_group IS NOT NULL AND going_group IS NOT NULL AND last_going_group <> going_group
The thread began with pulled-up-after-held-up (1.198 on 1329). It narrowed to jumps handicappers beaten 20+ lengths in big fields, where the never-involved flag carried the lift (1.324 on 1215). Adding a recent close finish lifted it to 1.72 on 236. H-332b49e9 now shows that most of this is a one-bad-run rebound. Jumps handicappers beaten 20+ lengths last time with a finish within 6 lengths in their last three runs score 1.232 on 1600 (p 0.004, periods 1.23 and 1.23, same-race control 0.98). Heavily beaten runners without a close finish score 1.023 on 5287. Outside the old flag/big-field niche the rule still scores 1.161 on 1359 (p 0.04), so the flag adds some lift but is not needed. Mark relief is not the carrier: runners not dropped in the weights score 1.307 on 506. Two warnings. First, removing recent winners (best_last3_beaten > 0) drops it to 1.152 on 931, and GB to 1.031. So the 669 recent winners carry about 1.34, which is the opposite of the parent niche, where removing winners raised it to 2.00. Second, the price profile is longshot-heavy. 16/1 or bigger scores 1.58 from only 22 wins, and SP ROI is +2.4% against BSP +37.9%. Below 16/1 it is 1.31 on 537 (up to 14/1) and 1.20 on 424 (up to 7/1), which is reassuring. All neighbours hold (1.16 to 1.26). Adding Flat races dilutes it: the implied Flat part is about 0.98 on about 1140, so the lift is jumps-only.
investigator v1
Answered17 Oct, 04:13
Is the jumps lift a rebound from one bad run after recent close form, holding without the never-involved flag and without a big last field?
Largely yes. The close-finish rule holds without the flag and big field at 1.161 on 1359 (p 0.04), and is 1.232 on 1600 overall, against 1.023 on 5287 with no close finish. The flag niche adds to it but is not needed.
perturbation: Perturb looser: 16+ lengths beaten, close finish within 7.5: A/E 1.11 on 2,355 bets, p = 0.061 (95%: 0.97 to 1.25), by period 1.10 · 1.12. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:14
perturbation: Perturb tighter: 24+ lengths beaten, close finish within 5: A/E 1.28 on 1,147 bets, p = 0.0035 (95%: 1.07 to 1.53), by period 1.28 · 1.29. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:14
H-332b49e9 (Rival: the lift is a one-bad-run rebound after close recent form, need) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:16from H-332b49e9
Tested17 Oct, 04:14
partition: Rival test at 14/1 or shorter: outside the parent niche: A/E 1.11 on 924 bets, p = 0.12 (95%: 0.93 to 1.33), by period 1.09 · 1.15. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:14
control: Control: close recent form but only a moderate beating last time (8-15 lengths): A/E 1.10 on 1,369 bets, p = 0.096 (95%: 0.95 to 1.26), by period 1.13 · 1.06. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:14
refinement: 16/1 or shorter, to check the lift is not a few longshot winners: A/E 1.21 on 650 bets, p = 0.042 (95%: 0.97 to 1.48), by period 1.28 · 1.06. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:14
refinement: excluding recent winners (close run was a defeat, not a win): A/E 1.11 on 523 bets, p = 0.24 (95%: 0.83 to 1.47), by period 1.04 · 1.26. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:14
H-c36bf81d (Test-mechanism: the heavy beating misleads most when ground or trip ch) passed discovery: A/E 1.26 on 917 bets (p 0.012), same-race control 0.98.
result0.60reasoned about 17 Oct, 04:16from H-c36bf81d
Tested17 Oct, 04:14
broad rule: A/E 1.34 on 669 bets, p = 0.0063 (95%: 1.07 to 1.65), by period 1.43 · 1.21. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:14
control: The key comparison is the same-going complement: the identical rule but with last_going_group = going_group. Th: A/E 1.20 on 682 bets, p = 0.074 (95%: 0.94 to 1.50), by period 1.10 · 1.26. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:14
refinement: Outside the never-involved niche, where the winner pattern reversed: A/E 1.28 on 343 bets, p = 0.077 (95%: 0.91 to 1.74), by period 1.14 · 1.47. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:14
control: Same-race rivals should score about 1.0, which shows the lift belongs to the runner and not to the kind of race: A/E 1.15 on 931 bets, p = 0.094 (95%: 0.93 to 1.41), by period 1.07 · 1.25. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 04:16
Does the recent-winner rebound hold below 16/1, or is the 1.34 carried by about 12 winners at 16/1 or bigger?
H-363890c3 scores 2.01 on 222 at 16/1 or bigger from 12 wins. The rest of the rule is only about 1.2, and the confidence interval runs from 1.07 to 1.65. Whether this is bettable depends on the band up to 14/1. Columns: sp_decimal, best_last3_beaten, last_beaten_lengths, race_type, is_handicap.
answeredpriority 0.60raised by O-c88bca5b
New question17 Oct, 04:16
Is the winner effect strongest when the heavy beating was the first run after the win: a bounce or a mark-hike run that the market reads as the horse being found out?
If the beating came straight after a win, the likely causes are a fresh mark rise, a bounce or a step up in grade, and those cluster in one run. A winner beaten 20+ lengths two runs after the win has more evidence of decline. Columns: days_since_win, days_since_last_run, last_days_since_last_run, last_official_rating, best_last3_beaten.
answeredpriority 0.50raised by O-c88bca5b, O-3ba2c8ff
New question17 Oct, 04:16
Why is the ground/trip-change rebound so much stronger in IRE (1.53 on 298) than GB (1.18 on 619)?
Irish jumps ground swings more, and IRE race_class is missing until 2026, so the GB/IRE gap could come from the going mix or from IRE price bands. A descriptive scan should cross region by last_going_group against going_group, and by sp_decimal band.
openpriority 0.30raised by O-b3d7bef8
New question17 Oct, 04:16
Is a change of conditions needed for the rebound, or does it work when the beating came on the same ground and at about the same trip?
raised with hypothesis 'Rival: the close-form rebound is just as strong when ground and trip are unchang'
To answer: Is the winner effect strongest when the heavy beating was the first run after the win: a bounce or a mark-hike run that the market reads as the horse being found out?
because A first run after a win is where one-off causes cluster: a bounce after a hard race, a fresh mark rise, a step up in grade. These are temporary, so the horse is close to the one that won. A horse that has run twice since winning has had more chance to show a real decline.
and the market misses it because The form line reads win and then 20+ lengths beaten, which looks like the handicapper has caught up with the horse. The market prices it as found out, when the beating may only be a bounce.
predicting positive, A/E 1.30 to 1.50 on 300+ bets, at or above the winner parent's 1.34
and it is wrong if A/E below 1.15, or no higher than winners whose win came two or more runs back
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND days_since_win IS NOT NULL AND last_days_since_last_run IS NOT NULL AND days_since_win = days_since_last_run + last_days_since_last_run
To answer: Is a change of conditions needed for the rebound, or does it work when the beating came on the same ground and at about the same trip?
because If the rebound only needs a single bad run that stands out against close form, it should hold without an excuse from ground or trip. Our estimate of about 1.19 for unchanged conditions, against 1.26 for a change, points that way.
and the market misses it because Without an obvious excuse for the beating, punters assume the horse has lost its form. So unchanged-conditions runners should be the least well understood if the rebound is real.
predicting positive, A/E 1.10 to 1.22 on 500+ bets, within 0.08 of the change group's 1.26
and it is wrong if A/E at or below 1.03, which would mean the ground/trip change is what carries the lift
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6 AND last_going_group IS NOT NULL AND last_going_group = going_group AND last_distance_yards IS NOT NULL AND distance_yards BETWEEN last_distance_yards - 300 AND last_distance_yards + 300
To answer: Does the recent-winner rebound hold below 16/1, or is the 1.34 carried by about 12 winners at 16/1 or bigger?
because If the rebound is real, a recent winner beaten 20+ lengths only once still has its ability. That should show at every price, including the shorter prices where most of the money goes. The rival is that 12 big-priced winners at 16/1 or bigger make the 1.34.
and the market misses it because The 20+ length beating is the most recent and most visible line of form. Punters and pricing models weight it more heavily than the earlier win, so even well-fancied winners drift past their true chance.
predicting positive, A/E 1.12 to 1.25 on 400+ bets at 14/1 or shorter
and it is wrong if A/E at or below 1.03 on 350+ bets at 14/1 or shorter, which would mean the winner lift is a longshot artefact
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten = 0 AND sp_decimal <= 15
The thread has moved away from the pulled-up idea it started from. It is now about jumps handicappers who were beaten 20+ lengths last time after recent form that was close. The general rule (H-332b49e9: beaten 20+ lengths last time, with a finish within 6 lengths in the last three runs) scored A/E 1.23 on 1600 and has now survived all 5 attacks. It is frozen for its vault query. Two child rules passed discovery:
- **Recent winners (H-363890c3):** 1.34 on 669 (p 0.006, control 0.97), the same in GB (1.33) and IRE (1.36). This confirms the figure we had worked out earlier: winners carry much of the lift. The weak point is where the lift sits. The 16/1-or-bigger band scores 2.01 on 222 bets from only 12 wins, so a handful of longshot winners could be inflating the headline.
- **Ground or trip change (H-c36bf81d):** 1.26 on 917, barely above the parent's 1.23. If we split the parent by bets, runners on similar conditions would score about 1.19 on about 680. So a change of conditions adds at most a little. The temporary-cause story (wrong trip or wrong ground explains the beating) is weakly supported. The rebound seems to happen whatever caused the bad run. The IRE part (1.53 on 298) is stronger than GB (1.18), and this group is also tilted towards 16/1 or bigger (1.73 on 309 from 14 wins).
What the chain has taught so far:
- A heavy beating that stands out against close recent form is mispriced over jumps, not on the Flat.
- Mark relief and class relief are not needed.
- The most open risk is that each sharpened version leans on a few big-priced winners.
The next tests should split the winner group by price and by what kind of run the beating was, not slice it further.
investigator v1
Answered17 Oct, 04:16
Among jumps handicappers beaten 20+ lengths, do recent winners (best_last3_beaten = 0) beat their price more than horses with a close placed finish, and why does this reverse the parent niche?
Yes on the first part. Recent winners beaten 20+ lengths score 1.34 on 669 (GB 1.33, IRE 1.36), above the roughly 1.15 for non-winning close finishers. Why it reverses the never-involved niche is still open, and 2.01 on 222 at 16/1 or bigger from 12 wins means part of it may be longshot noise.
Is the one-bad-run rebound strongest when the heavy beating came on different ground or at a different trip from today, as a temporary-cause mechanism predicts?
Mostly no. Runners whose beating came on different ground or at a different trip score 1.26 on 917 against the parent's 1.23, which leaves about 1.19 for unchanged conditions. A change adds little, so the rebound does not depend on the beating having an obvious temporary cause.
control: Same-going complement: identical rule with last_going_group = going_group: A/E 1.20 on 682 bets, p = 0.074 (95%: 0.94 to 1.50), by period 1.10 · 1.26. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
price_band: 16/1 or shorter only: is the lift a handful of longshot winners?: A/E 1.21 on 650 bets, p = 0.042 (95%: 0.97 to 1.48), by period 1.28 · 1.06. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
partition: Excluding recent winners: the close run was a defeat, not a win: A/E 1.11 on 523 bets, p = 0.24 (95%: 0.83 to 1.47), by period 1.04 · 1.26. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
partition: Quick return (60 days or less): going change without a season/yard-form shift: A/E 1.29 on 722 bets, p = 0.013 (95%: 1.03 to 1.59), by period 1.30 · 1.27. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:22
H-c36bf81d (Test-mechanism: the heavy beating misleads most when ground or trip ch) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:27from H-c36bf81d
Tested17 Oct, 04:22
perturbation: Perturb tighter: beaten 25+ last time, best of last 3 within 7.5 lengths: A/E 1.21 on 806 bets, p = 0.047 (95%: 0.97 to 1.50), by period 1.26 · 1.13. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
perturbation: Perturb looser: beaten 16+ last time, best of last 3 within 5 lengths: A/E 1.22 on 1,083 bets, p = 0.016 (95%: 1.02 to 1.45), by period 1.22 · 1.22. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:22
H-363890c3 (Sharpen: recent winners beaten 20+ lengths next time over jumps) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:27from H-363890c3
Tested17 Oct, 04:22
control: Control: placed but not won in last three (beaten >0 to 6 lengths): A/E 1.15 on 931 bets, p = 0.094 (95%: 0.93 to 1.41), by period 1.07 · 1.25. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
partition: Outside never-involved niche where the winner pattern reversed: A/E 1.28 on 343 bets, p = 0.077 (95%: 0.91 to 1.74), by period 1.14 · 1.47. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:22
H-b0d006d8 (Rival: the recent-winner rebound is a longshot artefact and vanishes a) passed discovery: A/E 1.26 on 447 bets (p 0.03), same-race control 0.97.
result0.60reasoned about 17 Oct, 04:27from H-b0d006d8
Tested17 Oct, 04:22
refinement: fields of 13 or fewer (field-size confounder): A/E 1.26 on 374 bets, p = 0.044 (95%: 0.97 to 1.61), by period 1.31 · 1.18. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
control: The other runners in the same races should come out near A/E 1.0 (about 0.95 to 1.02), which shows the lift bel: A/E 2.01 on 222 bets, p = 0.019 (95%: 1.04 to 3.51), by period 2.05 · 1.96. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
refinement: shorter than 16/1 (tests the longshot-concentration confounder): A/E 1.22 on 332 bets, p = 0.082 (95%: 0.92 to 1.60), by period 1.29 · 1.13. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:22
H-807e3cb2 (Sharpen: winners beaten 20+ lengths on their first run after winning) passed discovery: A/E 1.25 on 502 bets (p 0.049), same-race control 0.98.
result0.60reasoned about 17 Oct, 04:27from H-807e3cb2
Tested17 Oct, 04:22
refinement: quick turnaround after the win (bounce after a hard race): A/E 1.51 on 221 bets, p = 0.014 (95%: 1.04 to 2.11), by period 1.62 · 1.36. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:22
control: Same-race rivals should score around 1.0 (slightly below for the favourite-longshot pattern), so the lift belon: A/E 1.34 on 305 bets, p = 0.051 (95%: 0.94 to 1.85), by period 1.33 · 1.35. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:23
perturbation: Perturb: 16+ length beating with tighter 11/1 price cap: A/E 1.22 on 515 bets, p = 0.034 (95%: 0.98 to 1.50), by period 1.28 · 1.14. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:23
control: Control: same heavy beating with no close run in last three: A/E 1.01 on 2,552 bets, p = 0.41 (95%: 0.90 to 1.14), by period 1.06 · 0.92. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:23
H-b0d006d8 (Rival: the recent-winner rebound is a longshot artefact and vanishes a) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:27from H-b0d006d8
Tested17 Oct, 04:23
control: Control: win came two or more runs before the 20+ length beating: A/E 1.03 on 3,831 bets, p = 0.30 (95%: 0.92 to 1.16), by period 0.96 · 1.17. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:23
H-807e3cb2 (Sharpen: winners beaten 20+ lengths on their first run after winning) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:27from H-807e3cb2
Tested17 Oct, 04:23
partition: Quick turnaround after the win (bounce mechanism): A/E 1.51 on 221 bets, p = 0.014 (95%: 1.04 to 2.11), by period 1.62 · 1.36. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:23
H-8db9d29e (Rival: the close-form rebound is just as strong when ground and trip a) passed discovery as a niche: A/E 1.23 on 316 bets (p 0.132), same-race control 0.99.
result0.50reasoned about 17 Oct, 04:27from H-8db9d29e
Tested17 Oct, 04:23
control: Same-race rivals should score about 1.0. The main comparison is the change-of-conditions group (same close-form: A/E 1.23 on 1,283 bets, p = 0.0080 (95%: 1.04 to 1.45), by period 1.25 · 1.21. info
perturbation: Tighter cuts: 24L+ beating, close run within 5L, trip within 240y: A/E 1.36 on 189 bets, p = 0.098 (95%: 0.85 to 2.06), by period 0.99 · 1.52. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:24
perturbation: Looser cuts: 16L+ beating, close run within 7.5L, trip within 375y: A/E 1.06 on 516 bets, p = 0.35 (95%: 0.79 to 1.40), by period 0.96 · 1.12. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:24
H-8db9d29e (Rival: the close-form rebound is just as strong when ground and trip a) passed discovery but was falsified: driven by one price band (up to 14/1): the rest score 0.94.
near miss0.50reasoned about 17 Oct, 04:27from H-8db9d29e
Tested17 Oct, 04:24
control: Exclude the known never-involved 14+ field niche: A/E 1.11 on 269 bets, p = 0.32 (95%: 0.73 to 1.62), by period 1.09 · 1.13. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 04:27
Is there a dose-response in the margin: do recent winners beaten 8 to 19 lengths show a smaller lift than those beaten 20+?
If the market is misled by the size of the beating, a moderate beating should mislead less. If moderately beaten winners score as high, the edge is in recent winners being underbet in general, not in the beating. Columns: best_last3_beaten, last_beaten_lengths, days_since_win.
answeredpriority 0.45raised by O-ef2e2d1f, O-512837ca
New question17 Oct, 04:27
How much do the four frozen jumps rules (recent winner, sub-16/1 winner, first run after win, ground/trip change) share the same runners, and does each keep a lift on the runners the others miss?
Four frozen rules from one thread can look like replication when they are really one population. Before vault queries are spent, the lab needs to know whether the ground/trip-change rule adds runners beyond the recent-winner rules. Columns: best_last3_beaten, last_beaten_lengths, going_group, last_going_group, distance_yards, last_distance_yards, days_since_win.
testingpriority 0.55raised by O-1a2e3a25, O-f4de77f6, O-8e98be56, O-7cf74971
New question17 Oct, 04:27
Does the heavy-beating rebound need a visible excuse (ground or trip change) when there is no recent close finish?
The close-form rebound with ground and trip unchanged was killed as single-band driven (0.94 outside up to 14/1), while the ground/trip-change rule holds at 1.26 on 917. Testing the excuse without close form separates the two causes. Columns: going_group, last_going_group, distance_yards, last_distance_yards, best_last3_beaten.
answeredpriority 0.50raised by O-7450bbe8, O-d32706f6, O-f4de77f6
To answer: Is there a dose-response in the margin: do recent winners beaten 8 to 19 lengths show a smaller lift than those beaten 20+?
because A 20+ length beating looks to punters like the horse has been found out by its new mark. A moderate beating looks less damning, so the market should shade the price less and there is less mispricing to recover.
and the market misses it because The form figure and the large margin dominate how the horse reads on paper. A smaller margin reads as an ordinary defeat that is already roughly priced.
predicting positive, A/E 1.00 to 1.12 on 800+ bets, clearly below the 20+ group's 1.34
and it is wrong if A/E at or above 1.25, which would mean the margin is not the cause and recent winners are underbet after any defeat
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND best_last3_beaten = 0 AND last_position > 1 AND last_beaten_lengths >= 8 AND last_beaten_lengths < 20
To answer: Does the heavy-beating rebound need a visible excuse (ground or trip change) when there is no recent close finish?
because A heavy beating on different ground or at a different trip has an obvious excuse that the horse's recent form tends to confirm. Without recent close form, the beating may simply be true.
and the market misses it because Punters weight the last margin and pay little attention to whether the conditions of that run were against the horse.
predicting positive, A/E 1.08 to 1.18 on 600+ bets if the excuse alone matters
and it is wrong if A/E at or below 1.03, which would mean the ground/trip rule rides on close recent form
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten > 6 AND going_group IS NOT NULL AND last_going_group IS NOT NULL AND distance_yards IS NOT NULL AND last_distance_yards IS NOT NULL AND (going_group <> last_going_group OR distance_yards - last_distance_yards >= 440 OR last_distance_yards - distance_yards >= 440)
The thread has moved from 'pulled up after being held up' to a wider story: a jumps handicapper with good recent form that is beaten 20+ lengths gets underbet next time. Three results now back this. First, recent winners beaten 20+ lengths over jumps (H-363890c3) score 1.34 on 669. Second, its below-16/1 rival (H-b0d006d8) still scores 1.265 on 447 (p 0.03; up to 14/1: 1.335 on 207), so the lift is not just a longshot artefact. Third, the ground/trip-change version (H-c36bf81d) scores 1.26 on 917. All three survived every falsification attack and are frozen for the vault.
Restricting to the first run after the win (H-807e3cb2) did not sharpen anything. It fell to 1.25 on 502 with p 0.049 and a confidence interval reaching 0.96, and its 16/1+ band (1.54) rests on 7 wins. So the 'fresh mark rise or bounce' story is not where the effect concentrates.
The close-form rebound with ground and trip unchanged (H-8db9d29e) managed only 1.23 on 316 (p 0.13). It was killed because one band carried it: up to 14/1 scored 1.93 and the rest 0.94. That is a weak sign that a visible excuse (ground or trip change) is part of why the beating gets forgiven. These frozen rules overlap heavily, so they are not independent confirmations. Every number is discovery-only, with no forward bets.
investigator v1
Answered17 Oct, 04:27
Does the recent-winner rebound hold below 16/1, or is the 1.34 carried by about 12 winners at 16/1 or bigger?
Yes, it holds below 16/1: the sub-16/1 rival scores 1.265 on 447 (p 0.03), 1.335 on 207 up to 14/1, and it survived 6 of 6 attacks.
Is the winner effect strongest when the heavy beating was the first run after the win: a bounce or a mark-hike run that the market reads as the horse being found out?
No: requiring the heavy beating to be the first run after the win lowers A/E to 1.25 on 502 (p 0.049) from the parent's 1.34 on 669, so the effect is not concentrated in the bounce run.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:29
refinement: 14/1 or shorter (the earlier findings show the rebound holds below 16/1, while longshots inflate the broad A/E): A/E 1.16 on 677 bets, p = 0.062 (95%: 0.96 to 1.39), by period 1.13 · 1.20. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:29
refinement: GB only (checks the IRE-weighting confounder that drove the sister niche): A/E 1.12 on 607 bets, p = 0.13 (95%: 0.91 to 1.37), by period 1.11 · 1.14. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:29
refinement: Dose check: upper half of the band, 14 to 19 lengths (if margin drives the rebound, this half should beat the 8-13 half): A/E 0.94 on 323 bets, p = 0.67 (95%: 0.66 to 1.30), by period 0.78 · 1.15. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:29
H-32b815ff (Test-mechanism: recent winners beaten only 8 to 19 lengths should show) passed discovery as a niche: A/E 1.12 on 828 bets (p 0.117), same-race control 0.99.
result0.50reasoned about 17 Oct, 04:34from H-32b815ff
Tested17 Oct, 04:29
control: Other runners in the same races should show about 1.00 or a little below. If they do, any lift is not a propert: A/E 1.34 on 669 bets, p = 0.0063 (95%: 1.07 to 1.65), by period 1.43 · 1.21. info
refinement: Outside the never-involved big-field niche (overlap confounder): A/E 1.05 on 3,353 bets, p = 0.22 (95%: 0.92 to 1.20), by period 1.06 · 1.04. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:29
H-d586e8a0 (Rival: a ground or trip excuse alone, without recent close form, carri) passed discovery as a niche: A/E 1.05 on 4162 bets (p 0.214), same-race control 1.0.
result0.50reasoned about 17 Oct, 04:34from H-d586e8a0
Tested17 Oct, 04:29
control: Other runners in the same races should score about 1.0 (0.95 to 1.02); a lift there would mean a race property,: A/E 0.93 on 1,124 bets, p = 0.75 (95%: 0.72 to 1.18), by period 0.93 · 0.93. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:31
price_band: Price band: shorter than 8/1 only (excludes the 8/1-14/1 band at 1.625): A/E 1.08 on 458 bets, p = 0.24 (95%: 0.88 to 1.32), by period 1.05 · 1.11. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:31
H-32b815ff (Test-mechanism: recent winners beaten only 8 to 19 lengths should show) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:34from H-32b815ff
Tested17 Oct, 04:31
control: Control: same profile without a win in the last three runs: A/E 1.01 on 4,488 bets, p = 0.44 (95%: 0.92 to 1.10), by period 1.02 · 0.99. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:31
H-d586e8a0 (Rival: a ground or trip excuse alone, without recent close form, carri) passed discovery but was falsified: IRE A/E 0.937.
near miss0.50reasoned about 17 Oct, 04:34from H-d586e8a0
Tested17 Oct, 04:31
perturbation: Perturb looser: beaten 16+, close form > 5, trip gap 352y: A/E 1.02 on 5,206 bets, p = 0.34 (95%: 0.92 to 1.13), by period 1.02 · 1.02. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:31
perturbation: Perturb tighter: beaten 25+, close form > 7.5, trip gap 528y: A/E 1.11 on 3,093 bets, p = 0.080 (95%: 0.96 to 1.27), by period 1.11 · 1.10. pass
Why does IRE fail for the excuse-only rule (0.937) yet hold for the recent-winner rules (IRE 1.10)? Is Irish jumps pricing better at reading ground and trip excuses?
The IRE split of the excuse-only rule killed it, while the recent-winner rules hold in both countries. Comparing going_group against last_going_group and distance_yards against last_distance_yards by region would show whether Irish prices already contain the ground and trip change. Columns: region, going_group, last_going_group, distance_yards, last_distance_yards.
openpriority 0.25raised by O-b10f127d, O-a8e8d7c1
New question17 Oct, 04:34
Do recent winners priced 16/1 or bigger after a moderate (8 to 19 length) beating lose money consistently? (0.22 on 151 bets from 1 win)
This came out opposite to the parent's direction. Long prices after a recent win and a moderate defeat may mean the yard or the market knows the horse has gone off the boil. That is only a lead, since 151 bets from one win is thin. Columns: sp_decimal, last_beaten_lengths, best_last3_beaten, days_since_last_run, race_class, last_race_class.
openpriority 0.30raised by O-2ecb16ca
New question17 Oct, 04:34
Within the recent-winner jumps rebound, is the lift largest when the heavy beating came at a short last price, i.e. a fancied winner that flopped?
If the market overreacts to one bad number, a fancied flop (low last_sp_decimal) should mislead the most: the horse was expected to run well and the result hides that. If the lift is flat across last_sp_decimal, the edge is in recent winners being underbet in general. Columns: last_sp_decimal, last_beaten_lengths, best_last3_beaten, days_since_win, sp_decimal.
answeredpriority 0.45raised by O-2ecb16ca, O-4cd42568
To answer: Within the recent-winner jumps rebound, is the lift largest when the heavy beating came at a short last price, i.e. a fancied winner that flopped?
because A recent winner sent off fancied and then beaten 20+ lengths most likely had a one-off problem (ground, a mistake, a bleed, being eased once beaten). Its ability is unchanged, and the yard's confidence before that run was real information.
and the market misses it because The form line shows a big margin, and punters drift off a horse that 'flopped when fancied'. The dose-response (1.12 at 8 to 19 lengths, 1.34 at 20+) suggests the size of the margin drives the overreaction, and a fancied flop makes it look worse.
predicting positive, A/E 1.35 to 1.60 on 250+ bets
and it is wrong if A/E below 1.15, or no higher than the same rule with last_sp_decimal > 6
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten = 0 AND last_sp_decimal IS NOT NULL AND last_sp_decimal <= 6
To answer: How much do the four frozen jumps rules (recent winner, sub-16/1 winner, first run after win, ground/trip change) share the same runners, and does each keep a lift on the runners the others miss?
because A heavy beating straight after a win usually reflects one bad day, not lost ability. That holds whatever the race conditions, so novice and graded winners should rebound too.
and the market misses it because In non-handicaps the margin is the most visible form figure, and the win is easy to dismiss as a weaker race. That makes the price drift after a 20+ length defeat.
predicting positive, A/E 1.10 to 1.30 on 300+ bets
and it is wrong if A/E below 1.05, or the lift only at 16/1+
is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten = 0
The thread has narrowed from 'pulled up after being held up' to one claim: over jumps, a heavy beating misleads the market when the horse has shown real form very recently. These two observations sharpen that claim from both sides. First, the beating needs recent form. A ground or trip excuse on its own, with no recent close finish, scores only 1.05 on 4162 bets (p 0.21, CI 0.93 to 1.18). It was killed because IRE scored 0.937, while GB scored 1.098. So an excuse adds little without recent close form, and the ground/trip-change rule (1.26 on 917) was carried by its close-form condition. Second, the margin shows a weak dose-response. Recent winners beaten 8 to 19 lengths score 1.12 on 828 (p 0.12, control 0.99), against 1.34 on 669 for winners beaten 20+. Both GB (1.12) and IRE (1.10) agree. The rule survived 6 of 6 attacks, but its CI includes 1, so it is a weaker member of the same population, not a new edge. In that group the 16/1+ band scored 0.22 on 151 bets from a single win: moderately beaten winners priced as longshots look over-bet, not under-bet. Taken together: the bigger the gap between the recent win and the latest margin, the more the price lags. The lift comes from the horse's recent ability being masked by one bad number, not from a visible excuse. All of this is discovery-only. The frozen rules overlap heavily and none has forward bets.
investigator v1
Answered17 Oct, 04:34
Is there a dose-response in the margin: do recent winners beaten 8 to 19 lengths show a smaller lift than those beaten 20+?
Yes, weakly. Recent winners beaten 8 to 19 lengths score 1.12 on 828 (p 0.12) against 1.34 on 669 for those beaten 20+. That fits the size of the beating misleading the market, though the CIs overlap.
Does the heavy-beating rebound need a visible excuse (ground or trip change) when there is no recent close finish?
No. Without a recent close finish, a ground or trip change carries almost nothing: 1.05 on 4162 (p 0.21), with IRE at 0.937, and it was killed. The rebound needs recent close form.
H-8cff5453 (Replicate: recent winners beaten 20+ lengths in non-handicap jumps rac) was predicted positive but came out the other way: A/E 0.88 on 211 bets (p 0.268 for the opposite direction).
surprise0.87reasoned about 17 Oct, 04:37from H-8cff5453
Observed17 Oct, 04:36
H-40a72e24 (Sharpen: recent jumps winners whose 20+ length beating came at a short) passed discovery as a niche: A/E 1.18 on 226 bets (p 0.199), same-race control 0.99.
result0.50reasoned about 17 Oct, 04:37from H-40a72e24
Tested17 Oct, 04:36
broad rule (niche): A/E 1.18 on 226 bets, p = 0.20 (95%: 0.81 to 1.66), by period 1.06 · 1.32. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:36
control: Same-race other runners should be near 0.95-1.0, so the lift belongs to the selected horses and not to the race: A/E 1.46 on 443 bets, p = 0.0062 (95%: 1.09 to 1.91), by period 1.70 · 1.12. info
Replicate: recent winners beaten 20+ lengths in non-handicap jumps racesDiscovery → KILLEDfails: A/E 0.877, needs 1.05; p 0.7872, needs 0.05 or lower; periods [1.13, 0.63], each needs 1.0; A/E 0.877 against same-race control 1.013, needs 0.03 above (not a niche pass: A/E 0.877 below 1.05; p 0.7872054224138322 above 0.5; does not beat the same-race control 1.013; a period points the other way [1.13, 0.63])
Tested17 Oct, 04:36
neighbour: > 1 -> 0 (-20%): A/E 1.12 on 828 bets, p = 0.12 (95%: 0.93 to 1.34), by period 1.10 · 1.14. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 04:37
Is the non-handicap failure an IRE effect, where IRE non-handicap jumps are dominated by a few big yards whose beaten winners are correctly priced?
In the non-handicap replicate GB scored 1.15 on 81 and IRE 0.62 on 130. A descriptive look by region, trainer and is_novice within last_beaten_lengths >= 20 AND best_last3_beaten = 0 AND is_handicap = 0 would show whether the failure is broad or comes from IRE novice races. The samples are small, so expect a lead at best.
openpriority 0.30raised by O-6896f0b0
New question17 Oct, 04:37
Is the jumps recent-winner rebound handicap-only because the handicapper drops the mark after a heavy beating, so the lift comes from mark relief rather than from the market overreacting to form figures?
The non-handicap replicate came out the wrong way: 0.88 on 211, with IRE at 0.62 and GB at 1.15. In non-handicaps no mark is dropped, so a mark-relief mechanism would explain the gap. A test would split the handicap winner rule by rating_change (<0 against >=0), using official_rating and last_official_rating, both available from Nov 2025. If the lift is the same whether or not the mark fell, market overreaction stays the explanation.
answeredpriority 0.55raised by O-6896f0b0
New question17 Oct, 04:37
Is the short-last-price niche's 16/1+ lift (6 winners from 44) a real market overreaction to a fancied flop, or noise from a handful of winners?
The sharpen's whole edge is 4.83 on 44 bets at 16/1+, while up to 14/1 is 1.156 on 61. A descriptive check of last_sp_decimal against sp_decimal drift (the price lengthening from last time) across the wider heavy-beating close-form rule (best_last3_beaten <= 6) would show whether big drifters after a fancied flop beat their price in general, before any slice is tested.
To answer: Is the jumps recent-winner rebound handicap-only because the handicapper drops the mark after a heavy beating, so the lift comes from mark relief rather than from the market overreacting to form figures?
because A recent winner beaten 20+ lengths is often dropped a few pounds by the handicapper even though its ability has not changed. It runs back off a mark close to its winning one, which helps it in a handicap. A non-handicap has no mark to drop, and the replicate failed there (0.88).
and the market misses it because Punters read the big beating in the form figures and the small mark drop as minor, so they underrate a winner running off a mark close to the one it won off.
predicting positive, A/E 1.30 to 1.50 on 300+ bets, above the rating_change >= 0 complement by 0.10+
and it is wrong if A/E at or below 1.15, or no higher than the same rule with rating_change >= 0. Either would mean the mark drop does not carry the lift and market overreaction to the figures stays the explanation.
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten = 0 AND rating_change IS NOT NULL AND rating_change < 0
The thread's jumps rebound still rests on handicaps. Recent winners beaten 20+ lengths in jumps handicaps score 1.336 on 669 and hold under the 14/1 cap (1.265 on 447). The non-handicap replicate came out the opposite way to its prediction: 0.88 on 211, with IRE at 0.62 on 130 and GB at 1.15 on 81. So the mechanism is not simply that one bad number misleads the market everywhere. It may need the handicap structure: the mark falls after the beating and the horse runs back off a mark close to its winning one. That makes mark relief a real rival to market overreaction, and rating_change can test it. The fancied-flop sharpen did not sharpen anything: 1.18 on 226 (p 0.2), below the parent, and almost all of its lift is 6 winners at 16/1+ (4.83 on 44), with up to 14/1 at 1.156 on 61. That near miss should not be sliced further. Lesson so far: close recent form plus a heavy beating works in jumps handicaps only. Excuses and price-context cuts add nothing. The frozen rules overlap heavily and none has forward bets yet.
investigator v1
Answered17 Oct, 04:37
Within the recent-winner jumps rebound, is the lift largest when the heavy beating came at a short last price, i.e. a fancied winner that flopped?
No. Recent jumps winners whose heavy beating came at a short last price scored 1.18 on 226 (p 0.199), below the full recent-winner rule (1.336 on 669). A fancied flop does not mislead more, and what lift there is sits in 6 winners at 16/1+ (4.83 on 44).
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:41
perturbation: Perturb looser: last SP 7.0 or shorter, beaten 17+ lengths: A/E 1.31 on 357 bets, p = 0.032 (95%: 0.98 to 1.70), by period 1.27 · 1.35. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:41
perturbation: Perturb tighter: last SP 5.0 or shorter, beaten 24+ lengths: A/E 1.15 on 124 bets, p = 0.33 (95%: 0.66 to 1.86), by period 0.94 · 1.38. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:41
H-40a72e24 (Sharpen: recent jumps winners whose 20+ length beating came at a short) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:45from H-40a72e24
Tested17 Oct, 04:41
control: Control: fancied heavy flop with no recent close finish or win: A/E 1.08 on 645 bets, p = 0.27 (95%: 0.85 to 1.35), by period 1.11 · 1.00. pass
H-4392b730 (Rival: the jumps winner rebound is mark relief, strongest when the han) passed discovery as a niche: A/E 1.25 on 380 bets (p 0.08), same-race control 0.99.
result0.50reasoned about 17 Oct, 04:45from H-4392b730
Observed17 Oct, 04:41
H-4392b730 (Rival: the jumps winner rebound is mark relief, strongest when the han): the country 'GB' carries the effect (A/E 1.31) while the rest score IRE 0.96.
concentration0.55reasoned about 17 Oct, 04:45from H-4392b730
Tested17 Oct, 04:41
refinement: Price cap 14/1 or shorter (earlier work puts the rebound lift at mid prices, not longshots): A/E 1.20 on 251 bets, p = 0.15 (95%: 0.86 to 1.63), by period 1.30 · 1.03. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:41
refinement: Mark drop without a class drop (separates mark relief from easier company): A/E 1.28 on 180 bets, p = 0.16 (95%: 0.79 to 1.96), by period 1.61 · 0.69. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:41
control: The main comparison is the complement: the same recent winners beaten 20+ lengths whose mark was not dropped (r: A/E 1.41 on 269 bets, p = 0.031 (95%: 0.98 to 1.96), by period 1.56 · 1.27. info
Why did the price-band attack on H-40a72e24 count as held when it scored 1.004 on 182, below its stated kill line of 1.10?
A frozen rule whose sub-16/1 cut is at par, and whose unfancied control (1.455) beats it, is resting on 6 winners at 16/1+. The lab should check how attack verdicts are judged before spending a vault query. Columns: sp_decimal, last_sp_decimal.
openpriority 0.45raised by O-8e70e6ad
New question17 Oct, 04:45
Does an unchanged mark after a 20+ length beating mark a rebound in close-form jumps handicappers generally, beyond recent winners?
If the no-drop signal is real, it should lift the wider close-form rebound (best_last3_beaten <= 6, 1.232 on 1600) and not only winners. Columns: rating_change, best_last3_beaten, last_beaten_lengths, is_handicap, race_type.
answeredpriority 0.50raised by O-4822a69b
New question17 Oct, 04:45
Is the stronger rebound of undropped recent winners (rating_change >= 0) an official-excuse signal, or a quick return that came before the handicapper's reassessment?
The complement scored 1.412 on 269, against 1.249 for dropped horses, the opposite of the mark-relief prediction. If quick returners (days_since_last_run under about 10) carry it, it is a timing artefact. If horses returning after 14+ days with an unchanged mark carry it, the handicapper's refusal to drop is the information. Columns: rating_change, days_since_last_run, last_official_rating, official_rating, region.
answeredpriority 0.60raised by O-4822a69b, O-7693085d
To answer: Is the stronger rebound of undropped recent winners (rating_change >= 0) an official-excuse signal, or a quick return that came before the handicapper's reassessment?
because A yard that brings a recent winner back within about 10 days of a heavy beating knows the horse is well, and runs it before any rating change. The quick turnaround, not the handicapper's view, is the information.
and the market misses it because Punters see a 20-length beating a few days ago and assume the horse needs time. The yard's quick return signals that nothing is wrong.
predicting positive, A/E 1.30+ on 80 to 150 bets
and it is wrong if A/E at or below 1.10, or more than 0.10 below the days_since_last_run >= 14 version, which rules the quick-return explanation out
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten = 0 AND rating_change IS NOT NULL AND rating_change >= 0 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 10
To answer: Does an unchanged mark after a 20+ length beating mark a rebound in close-form jumps handicappers generally, beyond recent winners?
because For any horse with close recent form, a handicapper who holds its mark after a 20+ length beating has treated the run as an aberration. The close form stays the true guide.
and the market misses it because Punters weight the latest big margin and do not read the unchanged rating as an excuse.
predicting positive, A/E 1.20 to 1.35 on 500+ bets, above the rating_change < 0 version
and it is wrong if A/E at or below 1.10, or no higher than the same rule with rating_change < 0
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6 AND rating_change IS NOT NULL AND rating_change >= 0
To answer: Is the stronger rebound of undropped recent winners (rating_change >= 0) an official-excuse signal, or a quick return that came before the handicapper's reassessment?
because A handicapper who leaves a recent winner's mark alone after a 20+ length beating has judged that run as not true form (a mistake, wrong ground, eased). The horse's ability and winning mark stand, and it runs off a mark it has already proved it can win from.
and the market misses it because The form figures show a big beating, and punters read that and not the rating line. An unchanged mark looks like no relief, so it attracts no money, though it is the handicapper saying the run should be ignored.
predicting positive, A/E 1.30 to 1.50 on 180+ bets, at or above the quick-return (days_since_last_run <= 10) version
and it is wrong if A/E at or below 1.12, or below the same rule restricted to days_since_last_run <= 10, which would make the no-drop lift a timing artefact
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten = 0 AND rating_change >= 0 AND days_since_last_run >= 14
The mark-relief rival came out the wrong way, and that matters more than its niche pass. Recent jumps winners beaten 20+ lengths and then dropped in the weights scored 1.249 on 380 (p 0.08, GB 1.31, IRE 0.96). The same horses with their mark held or raised (rating_change >= 0) scored 1.412 on 269 (p 0.03, IRE 1.71 on 88). The prediction needed the dropped group to lead by 0.10, and it trails by 0.16. Mark relief is not what carries the winner rebound. The rebound is strongest when the handicapper saw the 20-length beating and did not move the mark, which reads like an official judgement that the run was not true form. The other explanation is a quick return before the reassessment took effect. The frozen fancied-flop rule (H-40a72e24) tells the same story from the other side. It survived 5 of 5 attacks, but its unfancied-last-run control scored 1.455 on 443, against 1.179 for the rule. Its ≤15/1 cut scored 1.004 on 182, below the attack's own kill line of 1.10, yet the attack is logged as held. Across the chain: the jumps handicap winner rebound is real in discovery (1.25 to 1.41 on every split). It is handicap-only, but not because of mark relief. Splitting it by the last price or by the mark drop moves the lift into the complement, not into the sharpened rule. That points to a market that overreacts to the visible beating, most where nothing else (a drop, a fancied price) flags it.
investigator v1
Answered17 Oct, 04:45
Is the jumps recent-winner rebound handicap-only because the handicapper drops the mark after a heavy beating, so the lift comes from mark relief rather than from the market overreacting to form figures?
No: mark relief does not carry it. Dropped winners scored 1.249 on 380, against 1.412 on 269 for those whose mark was held or raised. The handicap-only pattern needs another explanation, such as the handicapper's no-drop working as an excuse signal.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:47
H-4392b730 (Rival: the jumps winner rebound is mark relief, strongest when the han) passed discovery but was falsified: IRE A/E 0.96.
near miss0.50reasoned about 17 Oct, 04:52from H-4392b730
Tested17 Oct, 04:47
control: Control: mark dropped after a heavy beating, but no recent win: A/E 1.05 on 4,329 bets, p = 0.21 (95%: 0.94 to 1.17), by period 1.01 · 1.11. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:47
refinement: 15/1 or shorter (longshot confounder: 16/1+ scored 2.31 on 80): A/E 1.32 on 174 bets, p = 0.097 (95%: 0.87 to 1.92), by period 1.49 · 1.15. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:47
H-0b37c0ca (Rival: the no-drop lift is quick returners racing before the new mark ) pointed the predicted way but failed the bar: A/E 1.31 on 12 bets (p 0.452), control 0.96.
near miss0.60reasoned about 17 Oct, 04:52from H-0b37c0ca
Observed17 Oct, 04:47
H-6243bb06 (Test-mechanism: undropped recent winners after a heavy beating, given ) passed discovery: A/E 1.42 on 249 bets (p 0.034), same-race control 0.97.
result0.60reasoned about 17 Oct, 04:52from H-6243bb06
Tested17 Oct, 04:47
broad rule: A/E 1.31 on 12 bets, p = 0.45 (95%: 0.16 to 4.73). fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:47
refinement: beating came in a handicap (mark decision was about that run): A/E 1.44 on 221 bets, p = 0.034 (95%: 0.97 to 2.06), by period 1.59 · 1.31. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:47
control: Same-race other runners should sit near A/E 1.0 (about 0.95 to 1.0), showing the lift is not a property of thes: A/E 1.31 on 12 bets, p = 0.45 (95%: 0.16 to 4.73). info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:47
refinement: Outside the big-field never-involved niche (overlap with H-baef90d8): A/E 1.21 on 429 bets, p = 0.11 (95%: 0.89 to 1.60), by period 1.34 · 1.05. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:47
H-2af132dc (Generalise: an unchanged mark after a heavy beating lifts close-form j) passed discovery: A/E 1.31 on 506 bets (p 0.026), same-race control 0.98.
result0.60reasoned about 17 Oct, 04:52from H-2af132dc
Tested17 Oct, 04:47
control: Two comparisons. (1) Other runners in the same races should be about par (A/E about 0.95 to 1.0); if they are a: A/E 1.19 on 1,032 bets, p = 0.040 (95%: 0.98 to 1.43), by period 1.10 · 1.30. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:48
H-6243bb06 (Test-mechanism: undropped recent winners after a heavy beating, given ) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:52from H-6243bb06
Tested17 Oct, 04:48
control: Control: same horses but the mark WAS dropped: A/E 1.29 on 364 bets, p = 0.052 (95%: 0.95 to 1.73), by period 1.35 · 1.19. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:48
partition: No class drop today (rules out the class-relief confounder): A/E 1.33 on 134 bets, p = 0.15 (95%: 0.78 to 2.13), by period 1.26 · 1.44. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:48
partition: Beating came in a handicap (mark verdict was about that run): A/E 1.44 on 221 bets, p = 0.034 (95%: 0.97 to 2.06), by period 1.59 · 1.31. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:48
control: Close-form control: held mark after a heavy beating but no close recent form: A/E 0.83 on 735 bets, p = 0.89 (95%: 0.58 to 1.15), by period 0.99 · 0.49. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:48
perturbation: Tightened thresholds: 24L+ beating, close form within 5L: A/E 1.28 on 368 bets, p = 0.070 (95%: 0.92 to 1.73), by period 1.43 · 1.12. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 04:48
H-2af132dc (Generalise: an unchanged mark after a heavy beating lifts close-form j) survived falsification: 4 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 04:52from H-2af132dc
Tested17 Oct, 04:48
neighbour: >= 20 -> 16 (-20%): A/E 1.18 on 495 bets, p = 0.11 (95%: 0.90 to 1.51), by period 1.24 · 1.09. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 04:48
perturbation: Loosened thresholds: 16L+ beating, close form within 7.5L: A/E 1.16 on 724 bets, p = 0.11 (95%: 0.91 to 1.44), by period 1.29 · 1.01. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 04:52
How many of H-2af132dc's 506 runners are also in H-6243bb06, H-807e3cb2 and H-b0d006d8, and does the close-form non-winner remainder keep a lift on its own?
Six frozen rules from one thread can look like replication when they are one population. If the non-winner part of H-2af132dc (best_last3_beaten between 0 and 6, not 0) is about par, the generalisation is the winner rule repackaged. Columns: best_last3_beaten, rating_change, days_since_win, last_beaten_lengths.
answeredpriority 0.60raised by O-388a8a14, O-3761601e, O-74de1f0f
New question17 Oct, 04:52
Is the unchanged-mark rebound strongest when the in-running comment offers a visible excuse (eased, pulled up, jumping error), or when the beating had no visible excuse and only the handicapper's leniency marks it?
The figures pay where they disagree with what everyone saw. If the lift sits on beatings without a visible excuse, the handicapper's judgement is the private signal. If it sits on eased or error runs, the comment is public and the edge should be smaller. Columns: last_eased, last_jumping_error, jumping_errors_last3, last_never_involved, last_early_setback, rating_change.
openpriority 0.45raised by O-388a8a14, O-74de1f0f
New question17 Oct, 04:52
Within undropped close-form jumps runners, is the lift carried by marks left exactly unchanged (an excuse accepted) or by marks raised (a win penalty landing late, a horse rated better)?
'Undropped' joins two different stories. An unchanged mark means the handicapper excused the run. A raised mark usually means the earlier win's reassessment landed after the beating. Only the first fits the excuse mechanism. Columns: rating_change, last_official_rating, best_last3_beaten, last_beaten_lengths, days_since_win.
answeredpriority 0.55raised by O-388a8a14, O-b3313152
To answer: How many of H-2af132dc's 506 runners are also in H-6243bb06, H-807e3cb2 and H-b0d006d8, and does the close-form non-winner remainder keep a lift on its own?
because A handicapper who leaves a mark alone after a 20+ length beating judges the run to be unrepresentative. That judgement should apply to any horse with recent close form, not only to winners.
and the market misses it because Punters and form readers see the big margin last time and mark the horse down. The handicapper's decision not to drop the mark, which says the run was a one-off, is not read as a positive.
predicting positive, A/E 1.12 to 1.25 on 200+ bets
and it is wrong if A/E at or below 1.05, which would mean H-2af132dc rides on the recent-winner runners already in H-6243bb06
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten > 0 AND best_last3_beaten <= 6 AND rating_change IS NOT NULL AND rating_change >= 0
To answer: Within undropped close-form jumps runners, is the lift carried by marks left exactly unchanged (an excuse accepted) or by marks raised (a win penalty landing late, a horse rated better)?
because If excuse-reading is the cause, an exactly unchanged mark carries the lift. The rival: a raised mark means the horse's earlier win was reassessed upwards, so these are improvers the handicapper rates better, and the improvement, not the excuse, is the cause.
and the market misses it because Under the excuse story, the market sees the margin and not the handicapper's leniency. Under the rival, the market sees a higher mark and a bad last run and underrates a horse that is still improving.
predicting positive, A/E 1.20 to 1.40 on 300+ bets for rating_change = 0
and it is wrong if A/E at or below 1.10 on rating_change = 0, or the rating_change > 0 remainder scoring higher, which would favour the improver explanation
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6 AND rating_change IS NOT NULL AND rating_change = 0
The thread began with H36 (pulled up after being held up, 1.198 on 1329) and moved to jumps handicappers beaten 20+ lengths who had good recent form. Recent winners rebound, and the obvious explanation, mark relief, has now been ruled out twice. Dropped winners scored 1.249 on 380 against 1.412 on 269 for undropped ones. The mark-relief rival H-4392b730 passed discovery but was killed by IRE at 0.96. The unchanged-mark signal now generalises beyond winners. H-2af132dc covers close-form runners (best_last3 within 6 lengths) whose mark was not dropped after the heavy beating. It scored 1.31 on 506 (p 0.026, control 0.98), with GB 1.22 on 329 and IRE 1.55 on 177, and held 4 of 5 attacks. H-6243bb06 covers undropped recent winners who had time to be reassessed. It scored 1.42 on 249 (p 0.034, control 0.97) and held 6 of 6 attacks. The timing rival H-0b37c0ca (quick return before the new mark) found only 12 bets, so quick returners cannot carry the lift, whatever their A/E. The picture so far: when the handicapper sees a 20+ length beating and leaves the mark alone, that is official judgement that the run was not true form. The market reads the margin and lets the horse drift. The caution is overlap. At least six frozen rules in this thread share runners, and none has forward bets yet. They must not be counted as independent confirmations.
investigator v1
Answered17 Oct, 04:52
Does an unchanged mark after a 20+ length beating mark a rebound in close-form jumps handicappers generally, beyond recent winners?
Yes. An unchanged mark after a 20+ length beating lifts close-form jumps handicappers generally: 1.31 on 506 (p 0.026, control 0.98), positive in both GB (1.22) and IRE (1.55).
Is the stronger rebound of undropped recent winners (rating_change >= 0) an official-excuse signal, or a quick return that came before the handicapper's reassessment?
It is not a timing artefact. Only 12 undropped winners came back before reassessment, while horses given time to be reassessed scored 1.42 on 249 (H-6243bb06, 6/6 attacks held).
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:00
refinement: mark exactly unchanged (the 'held as a one-off' verdict; raised marks after a 20L beating mean something else and scored 0.57 among winners): A/E 1.28 on 169 bets, p = 0.16 (95%: 0.78 to 1.97), by period 1.44 · 1.01. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:00
refinement: 14/1 or shorter (story lifts have often rested on a few 16/1+ winners): A/E 1.27 on 158 bets, p = 0.14 (95%: 0.81 to 1.89), by period 1.38 · 1.13. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:00
refinement: 14+ days since last run (handicapper has had time to reassess, so the unchanged mark is a verdict and not just a delay): A/E 1.28 on 200 bets, p = 0.15 (95%: 0.81 to 1.92), by period 1.45 · 1.01. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 05:00
H-119fd2fb (Test-mechanism: the no-drop lift survives without recent winners (clos) passed discovery as a niche: A/E 1.18 on 237 bets (p 0.225), same-race control 1.0.
result0.50reasoned about 17 Oct, 05:10from H-119fd2fb
Observed17 Oct, 05:00
H-119fd2fb (Test-mechanism: the no-drop lift survives without recent winners (clos): the period '2025-10-01 to 2026-04-27' carries the effect (A/E 1.34) while the rest score after 2026-04-27 0.98.
concentration0.55reasoned about 17 Oct, 05:10from H-119fd2fb
Tested17 Oct, 05:00
control: (1) Other runners in the same races should be near par (about 0.95 to 1.0, as in the parent's 0.98). If they ar: A/E 1.15 on 652 bets, p = 0.14 (95%: 0.89 to 1.46), by period 0.95 · 1.39. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:00
refinement: GB only (IRE marks change less often, so check the unchanged-mark lift is not an IRE composition effect): A/E 1.33 on 251 bets, p = 0.058 (95%: 0.93 to 1.86), by period 1.46 · 1.15. info
H-4013e103 (Rival: the undropped lift is raised marks (late win penalties), not ma) passed discovery: A/E 1.44 on 392 bets (p 0.007), same-race control 0.96.
result0.60reasoned about 17 Oct, 05:10from H-4013e103
Tested17 Oct, 05:00
refinement: 15/1 or shorter (lift not resting on a few big-priced winners): A/E 1.44 on 277 bets, p = 0.010 (95%: 1.06 to 1.90), by period 1.57 · 1.26. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:00
control: The deciding comparison is the raised-mark remainder: the same close-form, 20+ length-beaten jumps handicappers: A/E 0.76 on 114 bets, p = 0.81 (95%: 0.31 to 1.57), by period 0.78 · 0.75. info
control: Control: same close-but-beaten heavy-beating group whose mark was dropped: A/E 1.15 on 652 bets, p = 0.14 (95%: 0.89 to 1.46), by period 0.95 · 1.39. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:01
perturbation: Perturb looser: beaten 16+ lengths, close form within 7.5 lengths: A/E 1.03 on 366 bets, p = 0.46 (95%: 0.71 to 1.44), by period 1.20 · 0.85. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 05:01
H-119fd2fb (Test-mechanism: the no-drop lift survives without recent winners (clos) passed discovery but was falsified: driven by one price band (up to 14/1): the rest score 0.96.
near miss0.50reasoned about 17 Oct, 05:10from H-119fd2fb
Tested17 Oct, 05:01
perturbation: Perturb tighter: beaten 25+ lengths, close form within 5 lengths: A/E 0.90 on 149 bets, p = 0.69 (95%: 0.46 to 1.56), by period 1.13 · 0.63. fail
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:01
price_band: Price band excluding 8/1 to 14/1: lift outside the one band carrying it: A/E 1.04 on 78 bets, p = 0.48 (95%: 0.57 to 1.75), by period 1.11. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:01
partition: Partition: exact unchanged mark after 14+ days (a real reassessment verdict): A/E 1.34 on 142 bets, p = 0.14 (95%: 0.79 to 2.11), by period 1.58 · 0.87. pass
perturbation: Perturb loose: 16+ lengths beaten, close form within 7.5 lengths: A/E 1.28 on 539 bets, p = 0.028 (95%: 0.99 to 1.63), by period 1.35 · 1.20. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:01
perturbation: Perturb tight: 25+ lengths beaten, close form within 5 lengths: A/E 1.51 on 262 bets, p = 0.013 (95%: 1.05 to 2.11), by period 1.67 · 1.33. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 05:01
H-4013e103 (Rival: the undropped lift is raised marks (late win penalties), not ma) survived falsification: 5 of 6 attacks left it standing; it is frozen for its one vault query.
result0.55reasoned about 17 Oct, 05:10from H-4013e103
Tested17 Oct, 05:01
partition: Partition: hurdles only: A/E 1.60 on 237 bets, p = 0.0080 (95%: 1.10 to 2.26), by period 1.93 · 1.21. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
New question17 Oct, 05:10
Does the unchanged-mark rebound after a heavy beating hold in Flat handicaps, where the plain close-form rebound failed (1.021 on 2055)?
H-4013e103's neighbour without the race_type filter scored 1.388 on 639. By subtraction, Flat handicaps contributed about 1.28 on about 247 bets. Because that number is inferred from a neighbour, it is only a lead. If the handicapper's held mark is the ingredient, it should work on the Flat even though close form alone did not. A test would read race_type, is_handicap, last_beaten_lengths, best_last3_beaten and rating_change.
testingpriority 0.40raised by O-b5e09f2e
New question17 Oct, 05:10
Is the fall from the first period to the second (1.58 to 1.28 unchanged; 1.34 to 0.98 for non-winners) a seasonal effect of summer jumping, or the edge fading?
Both periods sit inside one season. The second (after 2026-04-27) is summer jumping, with firmer ground and smaller fields. If the lift depends on going, the forward results this winter should come back up. If it is fading, forward bets will disappoint. A test would read going_group, last_going_group, field_size and the race date period.
openpriority 0.35raised by O-8445104c, O-b5e09f2e
New question17 Oct, 05:10
Is an unchanged mark after the heavy beating often mechanical, because the beating came in a non-handicap or under the other code's rating, so the handicapper never judged the run?
The excuse mechanism needs the handicapper to have looked at the run and declined to drop the mark. If the beating was in a non-handicap (last_is_handicap = 0), or in a chase when the horse runs today off its hurdle mark (last_race_type differs from race_type), rating_change = 0 tells us nothing about the handicapper's view. Splitting H-4013e103's 392 runners by last_is_handicap, and by last_race_type = race_type, decides whether the frozen rule's story is right. A test would read rating_change, last_is_handicap, last_race_type, race_type, last_official_rating and days_since_last_run.
testingpriority 0.60raised by O-b5e09f2e, O-b3b8f97d
To answer: Does the unchanged-mark rebound after a heavy beating hold in Flat handicaps, where the plain close-form rebound failed (1.021 on 2055)?
because If the active ingredient is the handicapper declining to drop a horse with recent close form after a heavy beating, the same verdict should mean the same thing on the Flat. The plain close-form rebound failed on the Flat (1.021), so if this passes, the held mark is what does the work, not close form.
and the market misses it because A Flat horse beaten 20+ lengths looks out of form on every form line. The unchanged mark is easy to miss, and bettors do not read it as the assessor's view that the run should be ignored.
predicting positive, A/E 1.15 to 1.35 on 200+ bets (the neighbour arithmetic suggests about 1.28 on about 247, so this is a lead, not a blind test)
and it is wrong if A/E at or below 1.05, or the lift existing only at 16/1+, which would mean the held-mark effect is specific to jumps
is_handicap = 1 AND race_type = 'Flat' AND last_beaten_lengths IS NOT NULL AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6 AND rating_change IS NOT NULL AND rating_change = 0
To answer: Is an unchanged mark after the heavy beating often mechanical, because the beating came in a non-handicap or under the other code's rating, so the handicapper never judged the run?
because Excuse story: after a handicap run, the official handicapper reviewed the 20+ length beating and declined to drop the mark, so the run was written off. Rival: when the beating came in a non-handicap, the mark often stays put by default. Then rating_change = 0 is just a record that nobody reassessed, and the lift comes from close form plus a heavy beating in a race the punters overrated. If the excuse story is right, the lift must hold when the last run was itself a handicap.
and the market misses it because The form line shows a 20+ length beating, and bettors mark the horse down. The rating line, showing that the handicapper reviewed the handicap run and kept the mark, is not read as a sign that the run was a one-off.
predicting positive, A/E 1.25 to 1.50 on 250+ bets with last_is_handicap = 1
and it is wrong if A/E at or below 1.10 on 200+ bets with last_is_handicap = 1, or the last_is_handicap = 0 remainder scoring higher, which would mean the unchanged mark is a non-reassessment artefact and not an excuse verdict
is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_is_handicap = 1 AND last_beaten_lengths >= 20 AND best_last3_beaten IS NOT NULL AND best_last3_beaten <= 6 AND rating_change IS NOT NULL AND rating_change = 0
The thread's latest split favours the excuse story over the improver rival. Close-form jumps handicappers beaten 20+ lengths whose mark was left exactly unchanged scored 1.444 on 392 bets (p 0.007, control 0.96). The rival group, same profile but mark raised, scored 0.762 on 114. A small drop of 1 to 3 lb scored 1.185 on 941. So the signal is the handicapper leaving the mark alone, not improvement. Note that H-4013e103's title says "raised marks" but its rule is rating_change = 0. H-4013e103 held 5 of 6 attacks and stayed stable at every neighbour (1.33 to 1.52), so it is frozen. Two weaknesses remain. GB alone is 1.335 (p 0.058) while IRE is 1.715, and the effect fell from 1.58 in the first period to 1.28 in the second. The close-form non-winner remainder (H-119fd2fb) was the test of whether the excuse effect reaches beyond recent winners. It mostly does not: 1.184 on 237, all of it in the 8/1 to 14/1 band (1.85 on 80), with 16/1+ at 0.44 and the later period at 0.98. It was killed. Its exactly-unchanged subset scored 1.278 on 169 (p 0.16, later period 1.01). By subtraction, the unchanged recent winners run about 1.57 on about 223 bets. The chain so far: close form plus a heavy beating gives a lift in jumps handicaps. Mark relief does not explain it. An unchanged mark sharpens it, and recent winners carry most of it. The rival not yet ruled out is that an 'unchanged' mark is often mechanical: the beating came in a non-handicap, so the handicapper never judged it at all. Everything rests on about one season of data.
investigator v1
Answered17 Oct, 05:10
How many of H-2af132dc's 506 runners are also in H-6243bb06, H-807e3cb2 and H-b0d006d8, and does the close-form non-winner remainder keep a lift on its own?
About 269 of H-2af132dc's 506 runners are recent winners. The close-form non-winner remainder scores 1.184 on 237, but only in the 8/1 to 14/1 band and at 0.98 in the later period, so it was killed and the generalisation rests on the winner part.
Within undropped close-form jumps runners, is the lift carried by marks left exactly unchanged (an excuse accepted) or by marks raised (a win penalty landing late, a horse rated better)?
Exactly unchanged marks carry the lift: rating_change = 0 scores 1.444 on 392 against 0.762 on 114 for raised marks, so the late-win-penalty/improver rival is ruled out.
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:11
refinement: GB only (marks reviewed more actively; tests the IRE confounder): A/E 1.36 on 228 bets, p = 0.052 (95%: 0.94 to 1.91), by period 1.51 · 1.16. info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 05:11
H-ef39e062 (Rival: the unchanged-mark lift is a mechanical non-reassessment, not t) passed discovery: A/E 1.48 on 350 bets (p 0.005), same-race control 0.96.
result0.60waiting for a reasoning passfrom H-ef39e062
Tested17 Oct, 05:11
control: Two comparisons. (1) The standard same-race control: the other runners in the same handicaps should score about: A/E 0.99 on 42 bets, p = 0.58 (95%: 0.20 to 2.90). info
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:13
partition: Review window: handicapper had time to reassess (8+ days since last run): A/E 1.53 on 337 bets, p = 0.0029 (95%: 1.14 to 2.02), by period 1.67 · 1.37. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:13
perturbation: Loosened thresholds: beaten 16+ lengths, close form within 7.5: A/E 1.27 on 476 bets, p = 0.045 (95%: 0.96 to 1.63), by period 1.36 · 1.16. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:13
perturbation: Tightened thresholds: beaten 24+ lengths, close form within 5: A/E 1.49 on 257 bets, p = 0.014 (95%: 1.05 to 2.07), by period 1.55 · 1.43. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Observed17 Oct, 05:13
H-ef39e062 (Rival: the unchanged-mark lift is a mechanical non-reassessment, not t) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query.
result0.55waiting for a reasoning passfrom H-ef39e062
Tested17 Oct, 05:13
broad rule (niche): A/E 1.28 on 247 bets, p = 0.15 (95%: 0.81 to 1.92), by period 1.20. pass
Bars grow from 1.00, the market. Faded bars have fewer than 40 bets.
Tested17 Oct, 05:13
refinement: not eased last time (beating was a real run, not a visible excuse): A/E 1.28 on 245 bets, p = 0.14 (95%: 0.81 to 1.92), by period 1.21. info
H-ade7c9c9 (Replicate: an unchanged mark after a 20+ length beating lifts close-fo): the country 'GB' carries the effect (A/E 1.35) while the rest score IRE 0.8.
concentration0.55waiting for a reasoning passfrom H-ade7c9c9
Observed17 Oct, 05:13
H-ade7c9c9 (Replicate: an unchanged mark after a 20+ length beating lifts close-fo) passed discovery as a niche: A/E 1.28 on 247 bets (p 0.146), same-race control 0.98.
result0.50waiting for a reasoning passfrom H-ade7c9c9
Tested17 Oct, 05:13
control: Same-race other runners should show about 1.0, so any lift belongs to the selected horses, not to the races. Al: A/E 0.88 on 807 bets, p = 0.82 (95%: 0.64 to 1.18), by period 0.90 · 0.87. info
