Raised easy handicap winners turned out within 14 days beat their price

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-97b1373c (Sharpen: easy handicap winners back within 14 days with no rise in the) pointed the predicted way but failed the bar: A/E 1.1 on 247 bets (p 0.208), control 0.95. O-d1dc9455
   • H-5c714267 (Test-mechanism: easy-winner quick-return lift lives in handicap-to-han) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query. O-e4c510dc
2. AskedWithin the handicap-to-handicap fortnight rule, is the lift carried by runners already raised by the handicapper (rating_change > 0), or by the rows where rating_change is missing? answered
3. MechanismThe handicapper raises the horse by its winning margin, and an easy win means the margin understated it, so the new mark still lags. A yard that brings it straight back is using a horse at its peak before the next rise and before its form fades.
4. Why the market misses itThe form book shows 'up 5lb for winning', a visible penalty that punters read as a reason to oppose. The 'won easily' comment that says the rise was too small is buried in the in-running notes.
5. Predictedpositive, A/E 1.20 to 1.35 on 150+ bets
6. Falsified ifA/E at or below the parent's 1.188, or below 1 in either period, or no better than 0.03 above the same-race control

Mechanism and rationale

The handicapper raises the horse by its winning margin, and an easy win means the margin understated it, so the new mark still lags. A yard that brings it straight back is using a horse at its peak before the next rise and before its form fades.

The rule

last_position = 1 AND last_won_easily = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change > 0

Reads: last_position, last_won_easily, days_since_last_run, is_handicap, last_is_handicap, rating_change

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

b9f9241a781bd07b0fbccc0e68ace54fc9c43b4419c0f6122ba41e821cc98e45

Frozen

16 Oct, 16:26

Nothing in the definition can change after this.

Confounders

• rating_change is NULL for some parent rows (missing ratings, first handicap-mark issues); the rule excludes them, so part of the implied ~1.28 complement may sit in NULLs and not in raised horses
• rating_change and last_won_easily only run from 2025-11, so the sample is short and from one season-and-a-bit; period splits will be thin
• Irish runners have thin coverage (IRE no-rise subset was 9 bets); any result is in effect GB-only
• the complement estimate (~231 bets, implied ~1.28) is arithmetic from two noisy numbers and may not repeat
• raised winners often go up in class too; class rises may confound the rating effect
• long-priced runners (16/1+) went 0 from 24 in the parent; the price mix of the raised subset may differ from the no-rise subset

Control

{
  "description": "Two comparisons. (1) The other runners in the same races should be at or below 1 (the parent's same-race control was 0.92); the rule must beat them by more than 0.03. (2) The question is whether the parent lift sits with already-raised horses, so the direct comparison is the same parent with no rise (rating_change <= 0). That subset already scored 1.096 on 247; the raised group must score clearly above it and above the parent's 1.188 for the mechanism to hold.",
  "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change <= 0"
}

Discovery result

pass broad rule · 16 Oct, 16:22

Bets

231

68 winners

A/E

1.33

95%: 1.04 to 1.69

SP return

+12.0%

level stakes

Betfair SP return

+30.3%

229 bets with a BSP

Strike rate 29.4%. p = 0.013. Same-race control: the other runners in these races had A/E 0.91, 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. 6 of 6 survived.

• perturbation: Tighter turnaround: back within 11 dayspass

  131 bets · A/E 1.18 (0.82 to 1.65) · control 0.95 · 16 Oct, 16:26

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27641.25

  after 2026-04-27671.12

• perturbation: Looser turnaround: back within 17 dayspass

  342 bets · A/E 1.32 (1.07 to 1.61) · control 0.91 · 16 Oct, 16:26

Sealed holdout (vault)

Not queried. Only a frozen idea can spend its one query against the 2020-25 holdout.

Forward record

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

  0 bets · · 16 Oct, 16:26

  Breakdown

  betsA/E0.71.01.3

  forward0–

Other experiments

• neighbour: <= 14 -> 11 (-20%)pass

  131 bets · A/E 1.18 (0.82 to 1.65) · control 0.95 · 16 Oct, 16:22

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27641.25

  after 2026-04-27671.12

• neighbour: <= 14 -> 17 (+20%)pass

  342 bets · A/E 1.32 (1.07 to 1.61) · control 0.91 · 16 Oct, 16:22

  Breakdown

Lineage

Agent runs

• designer v1ok · 40s · 16 Oct, 16:20

  Output (schema-checked)

  {
    "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change > 0",
    "refinements": [
      {
        "label": "16/1 or shorter",
        "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change > 0 AND sp_decimal <= 17"
      },
      {
        "label": "modest rise (1-7lb): the mark most likely still lags",
        "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change > 0 AND rating_change <= 7"
      }
    ],
    "control": {
      "description": "Two comparisons. (1) The other runners in the same races should be at or below 1 (the parent's same-race control was 0.92); the rule must beat them by more than 0.03. (2) The question is whether the parent lift sits with already-raised horses, so the direct comparison is the same parent with no rise (rating_change <= 0). That subset already scored 1.096 on 247; the raised group must score clearly above it and above the parent's 1.188 for the mechanism to hold.",
      "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change <= 0"
    },
    "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": [
      "rating_change is NULL for some parent rows (missing ratings, first handicap-mark issues); the rule excludes them, so part of the implied ~1.28 complement may sit in NULLs and not in raised horses",
      "rating_change and last_won_easily only run from 2025-11, so the sample is short and from one season-and-a-bit; period splits will be thin",
      "Irish runners have thin coverage (IRE no-rise subset was 9 bets); any result is in effect GB-only",
      "the complement estimate (~231 bets, implied ~1.28) is arithmetic from two noisy numbers and may not repeat",
      "raised winners often go up in class too; class rises may confound the rating effect",
      "long-priced runners (16/1+) went 0 from 24 in the parent; the price mix of the raised subset may differ from the no-rise subset"
    ],
    "mechanism": "After a handicap win the handicapper raises the horse by roughly its winning margin. An easy win means the margin understated its ability, so the new mark still lags its true level. A yard that brings it back within 14 days runs it at peak form before the next reassessment and before the form fades. So among quick-returning easy handicap winners, the lift should sit with those already carrying the rise (rating_change > 0) and not with those still running off the old mark.",
    "market_miss": "The form book shows 'up Xlb for winning', a visible penalty that punters read as a reason to oppose. The 'won easily' in-running comment, which says the rise was too small, is buried in the race notes and is not reflected in the price.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.20 to 1.35 on about 150-230 bets",
      "falsified_if": "A/E at or below the parent's 1.188, or below 1 in either period, or no more than 0.03 above the same-race control, or not above the no-rise control (rating_change <= 0)"
    },
    "notes": "The sketch is kept, with explicit NULL guards on days_since_last_run and rating_change. rating_change is pre_off and available from 2025-11 in both GB and IRE, the same window as last_won_easily, so no rows are lost to the window. The bet count may come in under 150 because NULL rating_change rows are excluded; a thin result pointing the right way goes on to falsification. The refinements come from the reasoning: the price cap from the parent's 0-from-24 at 16/1+, and the modest rise from the mechanism (a small rise after an easy win is the most likely to understate the horse)."
  }

Timeline

1. Job 1997 finished

   16 Oct, 16:26 · engine:compute-1

2. forward result for H-528222df (forward vs bar {'min_bets': 100, 'min_ae': 1.0, 'max_p': 0.2}): info

   16 Oct, 16:26 · engine:evaluator

3. compute-1 took job 1997 (forward_check, attempt 1)

   16 Oct, 16:26 · compute-1

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

   16 Oct, 16:26 · engine:scheduler

5. Job 1995 finished

   16 Oct, 16:26 · engine:compute-1

6. H-528222df Frozen → SHADOW vault skipped: last_won_easily, rating_change do not exist before the discovery zone; track it forward instead

   16 Oct, 16:26 · engine:orchestrator

7. compute-1 took job 1995 (vault, attempt 1)

   16 Oct, 16:26 · compute-1

8. Job 1994 finished

   16 Oct, 16:26 · engine:compute-1

9. Queued vault on the compute lane: frozen: one sealed vault query

   16 Oct, 16:26 · engine:scheduler

10. H-528222df Falsification → FROZEN survived falsification

    16 Oct, 16:26 · engine:falsification

11. falsification result for H-528222df (partition: Competitive fields: 8 or more runners): pass

    16 Oct, 16:26 · engine:evaluator

12. compute-1 took job 1994 (attack, attempt 1)

    16 Oct, 16:26 · compute-1

13. Job 1993 finished

    16 Oct, 16:26 · engine:compute-1

14. falsification result for H-528222df (partition: Flat handicaps only): pass

    16 Oct, 16:26 · engine:evaluator

15. compute-1 took job 1993 (attack, attempt 1)

    16 Oct, 16:26 · compute-1

16. Job 1992 finished

    16 Oct, 16:26 · engine:compute-1

17. falsification result for H-528222df (partition: No class rise: same or lower class than the win): pass

    16 Oct, 16:26 · engine:evaluator

18. compute-1 took job 1992 (attack, attempt 1)

    16 Oct, 16:26 · compute-1

19. Job 1991 finished

    16 Oct, 16:26 · engine:compute-1

20. falsification result for H-528222df (partition: Real rise only: raised 3lb or more): pass

    16 Oct, 16:26 · engine:evaluator