Rival: big-rise handicap winners beat their price at 15 to 35 days too, if the market over-fades the rise

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-dd3724ed (Test-mechanism: easy handicap winners raised 6 lb or more and back wit) passed discovery as a niche: A/E 1.47 on 131 bets (p 0.008), same-race control 0.85. O-46f43729
   • H-d9dabea2 (Rival: big-rise quick-returning handicap winners beat their price with) passed discovery as a niche: A/E 1.13 on 318 bets (p 0.16), same-race control 0.97. O-9d61b9a6
2. AskedDoes the market's over-fade of big handicap rises persist when the horse returns 15 to 35 days after the win, or is it tied to a quick turnout? testing
3. MechanismA rise of 6 lb or more marks real improvement. If punters read the headline rise as being found out by the handicapper, they under-price the horse whenever it next runs, not only within 14 days.
4. Why the market misses itThe form book shows a big jump in the official rating and the horse running off a career-high mark, which reads as handicapped out of it. The improvement that earned the rise is not in the number.
5. Predictedpositive, A/E 1.08 to 1.15 on 400+ bets if the over-fade is general; about 1.0 if the lift belongs to quick turnouts
6. Falsified ifA/E at or below 1.03 on 300+ bets, or below 1.0 in either period. That would tie the lift to the 14-day window, as with easy winners at 15 to 35 days (1.013 on 636).

Mechanism and rationale

A rise of 6 lb or more marks real improvement. If punters read the headline rise as being found out by the handicapper, they under-price the horse whenever it next runs, not only within 14 days.

The rule

last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6

Reads: last_position, days_since_last_run, is_handicap, last_is_handicap, rating_change

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

520c17270f207bf6f207f699b237eed38782af580fbc1adf0fec8a4228d1f950

Frozen

not frozen

The definition can still change.

Confounders

• Trainer intent: a horse given 15 to 35 days after a big rise may be freshened or aimed at a target race, so any lift may be the trainer's plan and not the market misreading the rise
• Class rise is more likely after a longer gap, so the horse may be running off the new mark in stronger company
• The Flat and jumps mix changes with the interval (jumpers return later), which could create or hide a lift
• rating_change only exists from November 2025 (about 11 months), so both periods are short and seasonal
• Winter breaks move the 15 to 35 day sample by season, mixing all-weather and jumps form
• Easy winners at 15 to 35 days scored only 1.013 on 636, so an easy-win share in this group drags the result toward 1.0 for reasons other than the rise

Control

{
  "description": "Same-race control: the other runners in the same handicaps should score about 0.90 to 1.0 A/E, as in the 14-day big-rise work (0.85). If the selection's lift is matched by its own race-mates, the effect is a property of these races, not of the rise. A second reference is the 14-day big-rise group (1.13 non-easy, 1.47 easy): the 15 to 35 day group should hold a lift if the market over-fades the rise, and be about 1.0 if the lift belongs to quick turnouts. This uses the existing measured groups and needs no new control rule.",
  "rule": null
}

Discovery result

fail broad rule · 16 Oct, 16:34

Bets

1,331

251 winners

A/E

1.00

95%: 0.88 to 1.13

SP return

-13.5%

level stakes

Betfair SP return

+3.2%

1,328 bets with a BSP

Strike rate 18.9%. p = 0.51. Same-race control: the other runners in these races had A/E 1.01, so the selection did not 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 · 37s · 16 Oct, 16:32

  Output (schema-checked)

  {
    "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6",
    "refinements": [
      {
        "label": "7/1 or shorter",
        "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND sp_decimal IS NOT NULL AND sp_decimal <= 8.0"
      },
      {
        "label": "Non-easy winners only (separates the rise from the easy-win flag, which failed at 15 to 35 days)",
        "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run BETWEEN 15 AND 35 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND last_won_easily = 0"
      }
    ],
    "control": {
      "description": "Same-race control: the other runners in the same handicaps should score about 0.90 to 1.0 A/E, as in the 14-day big-rise work (0.85). If the selection's lift is matched by its own race-mates, the effect is a property of these races, not of the rise. A second reference is the 14-day big-rise group (1.13 non-easy, 1.47 easy): the 15 to 35 day group should hold a lift if the market over-fades the rise, and be about 1.0 if the lift belongs to quick turnouts. This uses the existing measured groups and needs no new control rule.",
      "rule": null
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 150,
      "min_ae": 1.03,
      "max_p": 0.2
    },
    "confounders": [
      "Trainer intent: a horse given 15 to 35 days after a big rise may be freshened or aimed at a target race, so any lift may be the trainer's plan and not the market misreading the rise",
      "Class rise is more likely after a longer gap, so the horse may be running off the new mark in stronger company",
      "The Flat and jumps mix changes with the interval (jumpers return later), which could create or hide a lift",
      "rating_change only exists from November 2025 (about 11 months), so both periods are short and seasonal",
      "Winter breaks move the 15 to 35 day sample by season, mixing all-weather and jumps form",
      "Easy winners at 15 to 35 days scored only 1.013 on 636, so an easy-win share in this group drags the result toward 1.0 for reasons other than the rise"
    ],
    "mechanism": "A rise of 6 lb or more is evidence of real improvement. If punters read the headline rise as the handicapper having caught the horse, they under-price it whenever it next runs, not only when it comes back within 14 days. If so, the big-rise lift should hold at 15 to 35 days. If the lift comes from a quick turnout catching the horse at its peak, it should disappear at this interval.",
    "market_miss": "The form book shows a big jump in the official rating and a career-high mark, which reads as handicapped out of it. The improvement that earned the rise is not in that number. Punters anchor on the pounds added and fade the horse regardless of how long it has been off.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.08 to 1.15 on 400+ bets if the over-fade is general; about 1.0 if the lift belongs to quick turnouts",
      "falsified_if": "A/E at or below 1.03 on 300 or more bets, or A/E below 1.0 in either period. Either result ties the big-rise lift to the 14-day window, as with easy winners at 15 to 35 days (1.013 on 636)."
    },
    "notes": "I checked rating_change in the catalogue: it is pre-off and available from 2025-11-01 in both GB and IRE. I added NOT NULL guards to match the sibling rules. This is a rival-explanation test, so a null result near 1.0 is informative and supports the quick-turnout reading of H-dd3724ed and H-528222df. The 7/1-or-shorter refinement comes from the finding that big-rise quick returners do not win at 16/1 or bigger. The non-easy refinement makes the result comparable with the 1.13 non-easy 14-day group. Expect the IRE sample to be thin."
  }

Timeline

1. Job 2026 finished

   16 Oct, 16:34 · engine:compute-1

2. H-727481b2 Discovery → KILLED fails: A/E 0.999, needs 1.05; p 0.5142, needs 0.05 or lower; periods [1.04, 0.96], each needs 1.0; A/E 0.999 against same-race control 1.008, needs 0.03 above (not a niche pass: A/E 0.999 below 1.05; p 0.5142058585742307 above 0.5; does not beat the same-race control 1.008)

   16 Oct, 16:34 · engine:evaluator

3. discovery result for H-727481b2 (broad rule): fail

   16 Oct, 16:34 · engine:evaluator

4. H-727481b2 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:34 · engine:orchestrator

5. compute-1 took job 2026 (evaluate, attempt 1)

   16 Oct, 16:34 · compute-1

6. Job 2024 finished

   16 Oct, 16:32 · engine:ai-1

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

   16 Oct, 16:32 · engine:scheduler

8. H-727481b2 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:32 · agent:designer@v1

9. ai-1 took job 2024 (design, attempt 1)

   16 Oct, 16:32 · ai-1

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

    16 Oct, 16:32 · engine:scheduler

11. New idea H-727481b2: Rival: big-rise handicap winners beat their price at 15 to 35 days too, if the market over-fades the rise (rival-explanation)

    16 Oct, 16:32 · 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.