Test-mechanism: easy handicap winners still beat the price after reassessment (15-35 days)

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-5c714267 (Test-mechanism: easy-winner quick-return lift lives in handicap-to-han) passed discovery: A/E 1.19 on 478 bets (p 0.019), same-race control 0.92. O-dd247331
2. AskedDoes the easy-winner handicap lift last past the first reassessment (back in 15 to 35 days), or is it confined to runs before the handicapper has reacted? answered
3. MechanismIf the handicapper raises the mark only by the winning margin, an easy winner stays under-rated after reassessment too, so the lift should carry into runs at 15 to 35 days.
4. Why the market misses itPunters read the new mark and the bare margin, and both understate a horse that was eased or won on a tight rein.
5. Predictedpositive, A/E 1.08 to 1.15 on 400+ bets
6. Falsified ifA/E at or below 1.02, or below 1 in either period. That would put the lift in quick-return timing, not in mark lag.
7. Rule sketchlast_position = 1 AND last_won_easily = 1 AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1
8. question id

Mechanism and rationale

If the handicapper raises the mark only by the winning margin, an easy winner stays under-rated after reassessment too, so the lift should carry into runs at 15 to 35 days.

The rule

last_position = 1 AND last_won_easily = 1 AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1

Reads: last_position, last_won_easily, days_since_last_run, is_handicap, last_is_handicap

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

29b0e0aa8fc14b4b73194a7fed8ec705b6679e48143a5ebd71eaed6ef576ba39

Frozen

not frozen

The definition can still change.

Confounders

• Seasonal breaks in jumps campaigns put a different kind of horse into the 15-35 day window
• Trainers who wait out the reassessment differ from those who turn horses out quickly before the new mark applies
• The flat/jumps mix changes with gap length (flat handicappers return sooner)
• Price level: the parent's lift sat at short prices and 16/1+ went 0 from 24, so a different price mix in this window could hide or fake an effect
• Class rises after a win: a horse moving up a class after reassessment faces stronger rivals

Control

{
  "description": "Same-race other runners should sit near 0.92-1.0, as they did for the 14-day parent (control 0.92). The key comparison is last-time handicap winners that did NOT win easily, returning at 15-35 days into a handicap: if mark lag from an understated margin is the mechanism, these should be near 1.0 and clearly below the easy-winner rule. If they show the same lift, the effect is about recent handicap winners in general, not the easy-win label.",
  "rule": "last_position = 1 AND last_won_easily = 0 AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1"
}

Designer

designer@v1

Direction

positive

Discovery result

fail broad rule · 16 Oct, 16:19

Bets

636

126 winners

A/E

1.01

95%: 0.84 to 1.21

SP return

-11.4%

level stakes

Betfair SP return

+2.9%

635 bets with a BSP

Strike rate 19.8%. p = 0.45. Same-race control: the other runners in these races had A/E 1.00, so the selection beat its own races. Zone: discovery. Evaluator eval-1.0, data 6a4e98be18e83933

Falsification

No attacks yet. Only promising ideas are attacked.

Sealed holdout (vault)

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

Forward record

No forward record. Shadow tracking starts after the vault and counts only races run after the freeze.

Lineage

Agent runs

• designer v1ok · 33s · 16 Oct, 16:17

  Output (schema-checked)

  {
    "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1",
    "refinements": [
      {
        "label": "shorter than 16/1",
        "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1 AND sp_decimal < 17"
      },
      {
        "label": "flat only (cleaner reassessment cycle, removes jumps seasonal mix)",
        "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1 AND race_type = 'Flat'"
      }
    ],
    "control": {
      "description": "Same-race other runners should sit near 0.92-1.0, as they did for the 14-day parent (control 0.92). The key comparison is last-time handicap winners that did NOT win easily, returning at 15-35 days into a handicap: if mark lag from an understated margin is the mechanism, these should be near 1.0 and clearly below the easy-winner rule. If they show the same lift, the effect is about recent handicap winners in general, not the easy-win label.",
      "rule": "last_position = 1 AND last_won_easily = 0 AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1"
    },
    "pass_criteria": {
      "min_bets": 250,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Seasonal breaks in jumps campaigns put a different kind of horse into the 15-35 day window",
      "Trainers who wait out the reassessment differ from those who turn horses out quickly before the new mark applies",
      "The flat/jumps mix changes with gap length (flat handicappers return sooner)",
      "Price level: the parent's lift sat at short prices and 16/1+ went 0 from 24, so a different price mix in this window could hide or fake an effect",
      "Class rises after a win: a horse moving up a class after reassessment faces stronger rivals"
    ],
    "mechanism": "The handicapper reassesses mainly off the winning margin. A horse that was eased or won on a tight rein had more in hand than the margin shows, so its new mark still under-rates it. If so, the edge should survive the reassessment and show at 15-35 days, once the new mark applies, and not only in quick returns under the old mark (≤14 days, A/E 1.19 on 478).",
    "market_miss": "Punters price off the new official mark and the bare winning margin. Both understate a horse that won with plenty in hand. Once the rise is published, the market treats the horse as fairly handicapped.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.08 to 1.15 on 400+ bets",
      "falsified_if": "A/E at or below 1.02, or below 1.0 in either period, or no better than the non-easy last-time-winner control. Any of these would put the parent lift in quick-return timing before reassessment, not in mark lag."
    },
    "notes": "The rule is kept exactly as sketched. It uses only pre-off columns and mirrors the parent H-5c714267 with the day window moved from ≤14 to 15-35, so the comparison stays clean. min_bets is raised to 250 because the prediction expects 400+ bets. The refinements are taken from the findings: the lift sat at shorter prices and 16/1+ did nothing, and jumps seasonality is a named confounder. The broad rule is judged first."
  }

Timeline

1. Job 1967 finished

   16 Oct, 16:19 · engine:compute-1

2. H-4fd59906 Discovery → KILLED fails: A/E 1.013, needs 1.05; p 0.4521, needs 0.05 or lower; periods [1.14, 0.88], each needs 1.0; A/E 1.013 against same-race control 1.002, needs 0.03 above (not a niche pass: A/E 1.013 below 1.05; does not beat the same-race control 1.002; a period points the other way [1.14, 0.88])

   16 Oct, 16:19 · engine:evaluator

3. discovery result for H-4fd59906 (broad rule): fail

   16 Oct, 16:19 · engine:evaluator

4. H-4fd59906 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:19 · engine:orchestrator

5. compute-1 took job 1967 (evaluate, attempt 1)

   16 Oct, 16:19 · compute-1

6. Job 1965 finished

   16 Oct, 16:18 · engine:ai-1

7. Queued evaluate on the compute lane: discovery run of the broad rule

   16 Oct, 16:18 · engine:scheduler

8. H-4fd59906 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:18 · agent:designer@v1

9. ai-1 took job 1965 (design, attempt 1)

   16 Oct, 16:17 · ai-1

10. Queued design on the ai lane: write the pre-spec

    16 Oct, 16:17 · engine:scheduler

11. New idea H-4fd59906: Test-mechanism: easy handicap winners still beat the price after reassessment (15-35 days) (test-mechanism)

    16 Oct, 16:17 · agent:investigator@v1

Every change is an event with an actor: an agent and its version, an engine, a human, or the importer. Current stage: Killed.