Dose-response: turf-to-AW handicap switchers after a 61-149 day gap sit between the short- and long-gap lifts

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-e2d31d3d (Rival: turf-to-AW switchers after a short gap carry the same lift, so ) was predicted negative but came out the other way: A/E 1.07 on 4045 bets (p 0.089 for the opposite direction). O-c26a3db5
   • H-98ae1db9 (Turf-to-all-weather handicap returners after 150+ days beat their pric) passed discovery as a niche: A/E 1.23 on 654 bets (p 0.06), same-race control 0.97. O-e4c2a982
2. AskedWithin turf-to-AW Flat handicap switchers, does A/E rise steadily with the length of the layoff (14-60, 61-149, 150+ days), as a turf-discount-plus-readiness mechanism predicts? testing
3. MechanismThere are two discounts. Punters distrust turf form on the AW at any gap. On top of that they mark down rust, and the longer the absence the bigger the mark-down. But the trainer picks the return, so the extra rust discount is undeserved. If both parts are real, a 2-5 month gap should carry more lift than a 2-8 week gap and less than 150+ days.
4. Why the market misses itThe form book shows only turf runs and a long blank. Bettors weigh recent AW figures, and these horses have none. They treat the gap as a fitness worry rather than a planned return to a surface the yard chose.
5. Predictedpositive, A/E 1.08 to 1.18 on 800+ bets
6. Falsified ifA/E below 1.02 on 800+ bets. A flat middle would mean the 150+ day result is a lucky slice, not a dose-response.

Mechanism and rationale

There are two discounts. Punters distrust turf form on the AW at any gap. On top of that they mark down rust, and the longer the absence the bigger the mark-down. But the trainer picks the return, so the extra rust discount is undeserved. If both parts are real, a 2-5 month gap should carry more lift than a 2-8 week gap and less than 150+ days.

The rule

race_type = 'Flat' AND is_handicap = 1 AND surface = 'AW' AND last_surface = 'Turf' AND days_since_last_run BETWEEN 61 AND 149

Reads: race_type, is_handicap, surface, last_surface, days_since_last_run

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

6c673b31e5e010a24e360704a2918e8c2bfde2d0a8d81892598d498e13eb9c15

Frozen

not frozen

The definition can still change.

Confounders

• Overlap with the 120+ day neighbour (1.295 on 899): the 120-149 day slice is shared, so the 61-119 refinement is the independent test of the middle dose
• Calendar mix: mid-gap switchers are mostly late-autumn and winter AW runners coming off a turf season, while 150+ day returners are often spring returners. Season, not gap length, may set the lift
• Ireland: its AW programme is small (Dundalk) and has underperformed in every AW rule in this thread
• Longshot concentration: earlier switcher lifts leaned on a few 16/1+ winners, so a pooled A/E can rest on a handful of results
• Period decay: later periods have been weaker in every rule, so the market may be fading the edge
• Short-gap switchers already score 1.07, so a middle result near 1.07 cannot tell dose-response apart from a flat surface-switch effect

Control

{
  "description": "Other runners in the same races should score about 0.97 to 1.00, so the lift belongs to the switchers and not to the races. Separate control: AW-to-AW Flat handicap runners after the same 61-149 day gap. These carry the rust discount but not the surface-distrust discount, and they should be close to fairly priced (about 0.98 to 1.03, like the 1.015 seen for AW-to-AW 150+ day returners). If the control also shows the lift, the gap is driving it and the surface switch is not.",
  "rule": "race_type = 'Flat' AND is_handicap = 1 AND surface = 'AW' AND last_surface = 'AW' AND days_since_last_run BETWEEN 61 AND 149"
}

Discovery result

fail broad rule · 16 Oct, 14:13

Bets

1,036

81 winners

A/E

0.99

95%: 0.79 to 1.23

SP return

-26.8%

level stakes

Betfair SP return

+8.5%

989 bets with a BSP

Strike rate 7.8%. p = 0.54. 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 · 31s · 16 Oct, 14:07

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND is_handicap = 1 AND surface = 'AW' AND last_surface = 'Turf' AND days_since_last_run BETWEEN 61 AND 149",
    "refinements": [
      {
        "label": "GB only",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND surface = 'AW' AND last_surface = 'Turf' AND days_since_last_run BETWEEN 61 AND 149 AND region = 'GB'"
      },
      {
        "label": "14/1 or shorter (no longshot reliance)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND surface = 'AW' AND last_surface = 'Turf' AND days_since_last_run BETWEEN 61 AND 149 AND sp_decimal <= 15"
      },
      {
        "label": "61-119 days (independent of the 120+ day neighbour)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND surface = 'AW' AND last_surface = 'Turf' AND days_since_last_run BETWEEN 61 AND 119"
      }
    ],
    "control": {
      "description": "Other runners in the same races should score about 0.97 to 1.00, so the lift belongs to the switchers and not to the races. Separate control: AW-to-AW Flat handicap runners after the same 61-149 day gap. These carry the rust discount but not the surface-distrust discount, and they should be close to fairly priced (about 0.98 to 1.03, like the 1.015 seen for AW-to-AW 150+ day returners). If the control also shows the lift, the gap is driving it and the surface switch is not.",
      "rule": "race_type = 'Flat' AND is_handicap = 1 AND surface = 'AW' AND last_surface = 'AW' AND days_since_last_run BETWEEN 61 AND 149"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 150,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Overlap with the 120+ day neighbour (1.295 on 899): the 120-149 day slice is shared, so the 61-119 refinement is the independent test of the middle dose",
      "Calendar mix: mid-gap switchers are mostly late-autumn and winter AW runners coming off a turf season, while 150+ day returners are often spring returners. Season, not gap length, may set the lift",
      "Ireland: its AW programme is small (Dundalk) and has underperformed in every AW rule in this thread",
      "Longshot concentration: earlier switcher lifts leaned on a few 16/1+ winners, so a pooled A/E can rest on a handful of results",
      "Period decay: later periods have been weaker in every rule, so the market may be fading the edge",
      "Short-gap switchers already score 1.07, so a middle result near 1.07 cannot tell dose-response apart from a flat surface-switch effect"
    ],
    "mechanism": "There are two separate discounts on turf-to-AW handicap switchers. (1) Surface distrust: punters mark down turf-only form on the AW at any gap. Short-gap switchers show this at about 1.07. (2) A rust discount that grows with the absence, even though the trainer chose when to bring the horse back and onto which surface, so the extra mark-down is undeserved. If both are real and the second scales with the gap, 61-149 day switchers should land between the short-gap lift (about 1.07) and the 150+ day lift (about 1.23).",
    "market_miss": "The form book shows only turf runs and then a blank of two to five months. Bettors rely on recent AW figures, and these horses have none. They read the gap as a fitness worry rather than a planned return to a surface the yard chose, so they discount the horse twice: once for the surface and once for the absence.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.08 to 1.18 on 800+ bets, between the short-gap 1.07 and the 150+ day 1.23",
      "falsified_if": "A/E below 1.02 on 800+ bets (a flat middle, so the 150+ day result is a lucky slice and not a dose-response), or a lift that lives only in 120-149 days with 61-119 at or below 1.02, or the AW-to-AW same-gap control scoring as well as the switchers."
    },
    "notes": "I kept the rule sketch as it was: every column it uses is pre-off and the band is closed (61-149), so it neither overlaps the short-gap range (14-60) nor the 150+ day range. The refinements come straight from the listed confounders: Ireland, the longshot tail, and the overlap with the 120+ day neighbour. I raised min_bets to 300 because the prediction expects 800+ bets. Judge dose-response by the ordering short < mid < long, not by this rule alone."
  }

Timeline

1. Job 1751 finished

   16 Oct, 14:13 · engine:compute-1

2. H-f00f8c45 Discovery → KILLED fails: A/E 0.992, needs 1.05; p 0.5445, needs 0.05 or lower; periods [0.99, 1.0], each needs 1.0; A/E 0.992 against same-race control 1.006, needs 0.03 above (not a niche pass: A/E 0.992 below 1.05; p 0.5444711738961832 above 0.5; does not beat the same-race control 1.006)

   16 Oct, 14:13 · engine:evaluator

3. discovery result for H-f00f8c45 (broad rule): fail

   16 Oct, 14:13 · engine:evaluator

4. H-f00f8c45 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 14:13 · engine:orchestrator

5. compute-1 took job 1751 (evaluate, attempt 1)

   16 Oct, 14:13 · compute-1

6. Job 1749 finished

   16 Oct, 14:07 · engine:ai-1

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

   16 Oct, 14:07 · engine:scheduler

8. H-f00f8c45 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 14:07 · agent:designer@v1

9. ai-1 took job 1749 (design, attempt 1)

   16 Oct, 14:07 · ai-1

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

    16 Oct, 14:07 · engine:scheduler

11. New idea H-f00f8c45: Dose-response: turf-to-AW handicap switchers after a 61-149 day gap sit between the short- and long-gap lifts (test-mechanism)

    16 Oct, 14:07 · 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.