Lightly raced four-year-old returners on the all-weather: is the edge really turf-only?

Reasoning

Why this hypothesis exists: what the lab saw, the question it asked, the mechanism it proposed, what it predicted, and the result that would prove it wrong. Written before the test was run.

1. Observed
   • H-29eb96fe (Lightly raced four-year-olds back from the winter on turf only) survived falsification: 5 of 6 attacks left it standing; it is frozen for its one vault query. O-7e9098fb
   • H-29eb96fe (Lightly raced four-year-olds back from the winter on turf only) passed discovery as a niche: A/E 1.2 on 379 bets (p 0.127), same-race control 0.97. O-94a1c4ce
   • H-29eb96fe (Lightly raced four-year-olds back from the winter on turf only) met the sealed vault supportive: A/E 1.1 on 1741 bets (p 0.102, threshold 0.0033). O-bbb9c77c
2. AskedIs the H126 lift and its second-run follow-up a late-spring turf effect, weak or absent on the all-weather in March and early April? answered
3. MechanismIf the cause is maturation plus the mark, it should not depend on surface. If it is going or late-spring turf readiness, all-weather returners should score about 1.0.
4. Why the market misses itEarly-season all-weather handicaps are dominated by fit, race-hardened horses. The market prices fresh returners against that sharpness and may over-discount the winter's physical gain.
5. Predictedpositive, A/E 1.05 to 1.12 on 400+ bets
6. Falsified ifA/E below 1.00 on 400+ bets, which would favour the turf-only explanation and argue for restricting the system

Mechanism and rationale

If the cause is maturation plus the mark, it should not depend on surface. If it is going or late-spring turf readiness, all-weather returners should score about 1.0.

The rule

race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND surface = 'AW'

Reads: race_type, is_handicap, month, age, prior_runs, days_since_last_run, surface

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

7447c0cc1da81e618f6153b46a6d13e06f465e58aaf7b0b058b2de341956b778

Frozen

16 Oct, 12:29

Nothing in the definition can change after this.

Confounders

• Calendar: AW returners run mostly in March and early April, the earlier part of the spring, so surface and season-stage are confounded (refinement 2 addresses this)
• Only one spring of post-October-2025 data, so surface and period are confounded in any period split
• Race class: AW spring handicaps tend to be lower class than turf ones, so field quality differs
• Trainer selection: yards that choose an AW comeback may differ in intent and readiness from those waiting for turf
• Longshot artefact: earlier lift in this thread sat in the up-to-14/1 band, so a small number of big-priced winners could create or hide an effect (refinement 1)

Control

{
  "description": "Same-race other runners: expected about 1.00, since early-season AW handicaps carry no general mispricing. If they also show lift, the effect is a property of these races, not of the returners. Separate control rule: older (5+) handicappers returning on the AW after 150+ days in the same months. If they show the same lift, the effect is about freshness on the AW in general, not four-year-old maturation. Expected about 1.00 or below.",
  "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age >= 5 AND days_since_last_run >= 150 AND surface = 'AW'"
}

Designer

Discovery result

pass broad rule (niche) · 16 Oct, 12:28

Bets

140

19 winners

A/E

1.18

95%: 0.71 to 1.84

SP return

-29.9%

level stakes

Betfair SP return

-18.9%

140 bets with a BSP

Strike rate 13.6%. p = 0.27. Same-race control: the other runners in these races had A/E 0.97, so the selection beat its own races. Zone: discovery. Evaluator eval-1.0, data 6a4e98be18e83933

Falsification

Each attack is a deliberate attempt to kill the idea. 4 of 5 survived.

• perturbation: Loosen thresholds: up to 12 runs, 120+ days offpass

  195 bets · A/E 1.03 (0.65 to 1.56) · control 0.99 · 16 Oct, 12:29

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-271621.12

  after 2026-04-2733–

• perturbation: Tighten thresholds: up to 8 runs, 180+ days offpass

  75 bets · A/E 1.22 (0.59 to 2.25) · control 0.97 · 16 Oct, 12:29

  Breakdown

Sealed holdout (vault)

Bets

698

16 Oct, 12:29

A/E

1.10

p = 0.20

BH threshold then

0.0032

over every vault test so far

Outcome

Failed

Forward record

• forward vs bar {'min_bets': 100, 'min_ae': 1.0, 'max_p': 0.2}info

  0 bets · · 16 Oct, 12:29

  Breakdown

  betsA/E0.71.01.3

  forward0–

Other experiments

• neighbour: <= 10 -> 8 (-20%)pass

  109 bets · A/E 1.24 (0.71 to 2.01) · control 0.96 · 16 Oct, 12:58

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27871.41

  after 2026-04-2722–

• neighbour: <= 10 -> 12 (+20%)pass

  161 bets · A/E 1.09 (0.67 to 1.69) · control 0.98 · 16 Oct, 12:58

  Breakdown

Lineage

Agent runs

• designer v1ok · 30s · 16 Oct, 12:24

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND surface = 'AW'",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND surface = 'AW' AND sp_decimal <= 15"
      },
      {
        "label": "April-May only (calendar overlap with the turf season)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND surface = 'AW'"
      }
    ],
    "control": {
      "description": "Same-race other runners: expected about 1.00, since early-season AW handicaps carry no general mispricing. If they also show lift, the effect is a property of these races, not of the returners. Separate control rule: older (5+) handicappers returning on the AW after 150+ days in the same months. If they show the same lift, the effect is about freshness on the AW in general, not four-year-old maturation. Expected about 1.00 or below.",
      "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age >= 5 AND days_since_last_run >= 150 AND surface = 'AW'"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Calendar: AW returners run mostly in March and early April, the earlier part of the spring, so surface and season-stage are confounded (refinement 2 addresses this)",
      "Only one spring of post-October-2025 data, so surface and period are confounded in any period split",
      "Race class: AW spring handicaps tend to be lower class than turf ones, so field quality differs",
      "Trainer selection: yards that choose an AW comeback may differ in intent and readiness from those waiting for turf",
      "Longshot artefact: earlier lift in this thread sat in the up-to-14/1 band, so a small number of big-priced winners could create or hide an effect (refinement 1)"
    ],
    "mechanism": "Four-year-olds with 10 or fewer runs who have had a winter off often come back physically stronger than their handicap mark, which is still based on their three-year-old form. If maturation plus a lagging mark is the cause, the effect should not depend on surface, so AW returners should beat their prices as turf returners do (H-29eb96fe: 1.20 on 379). If the turf-only result reflects going or late-spring turf readiness, AW returners should score about 1.00.",
    "market_miss": "Early-season AW handicaps are dominated by fit, race-hardened horses that ran through the winter. The market prices a fresh returner against that sharpness and marks it down for its absence, which may over-discount the physical gain a lightly raced horse made between three and four.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.05 to 1.12 on 400+ bets, a bit below the turf subset",
      "falsified_if": "A/E below 1.00 on 400+ bets (or below 1.00 with the same-race control at about 1.00). That would favour the turf-only explanation and argue for restricting the live system to turf."
    },
    "notes": "The rule sketch's surface <> 'Turf' was changed to surface = 'AW'. The catalogue has only these two surface values, so the rule is the same in substance, but it is explicit and leaves out any null or unknown surfaces. Otherwise this is H126 with the opposite surface to H-29eb96fe, so the two subsets add up to the full H126 rule (excluding any null surfaces). A result near 1.00 here is informative rather than a failure of the lab's thread: it supports restricting SYS-four-year-old-returns-after-a-win to turf. Refinement 2 drops March to test whether the AW result changes once its calendar overlaps with the turf subset."
  }

Timeline

1. Job 1518 finished

   16 Oct, 12:58 · engine:compute-1

2. hypothesis note H-b05a33c3

   16 Oct, 12:58 · engine:perturb

3. perturbation result for H-b05a33c3 (neighbour: drop: is_handicap = 1): pass

   16 Oct, 12:58 · engine:perturb

4. perturbation result for H-b05a33c3 (neighbour: drop: race_type = 'Flat'): pass

   16 Oct, 12:58 · engine:perturb

5. perturbation result for H-b05a33c3 (neighbour: >= 150 -> 180 (+20%)): pass

   16 Oct, 12:58 · engine:perturb

6. perturbation result for H-b05a33c3 (neighbour: >= 150 -> 120 (-20%)): pass

   16 Oct, 12:58 · engine:perturb

7. perturbation result for H-b05a33c3 (neighbour: <= 10 -> 12 (+20%)): pass

   16 Oct, 12:58 · engine:perturb

8. perturbation result for H-b05a33c3 (neighbour: <= 10 -> 8 (-20%)): pass

   16 Oct, 12:58 · engine:perturb

9. compute-1 took job 1518 (mutate, attempt 1)

   16 Oct, 12:58 · compute-1

10. Queued mutate on the compute lane: parameter stability: do neighbouring thresholds hold?

    16 Oct, 12:58 · engine:scheduler

11. Job 1359 finished

    16 Oct, 12:29 · engine:compute-1

12. forward result for H-b05a33c3 (forward vs bar {'min_bets': 100, 'min_ae': 1.0, 'max_p': 0.2}): info

    16 Oct, 12:29 · engine:evaluator

13. compute-1 took job 1359 (forward_check, attempt 1)

    16 Oct, 12:29 · compute-1

14. Queued forward_check on the compute lane: daily forward check of a shadow system

    16 Oct, 12:29 · engine:scheduler

15. Job 1354 finished

    16 Oct, 12:29 · engine:compute-1

16. H-b05a33c3 Vault → SHADOW vault SUPPORTIVE: A/E 1.101 on 698 bets, p 0.2014 against BH threshold 0.003226 over 31 vault tests

    16 Oct, 12:29 · engine:orchestrator