Test: the trip-drop lift is absent on Flat all-weather

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-cde96a9f (Rival: the trip-drop lift is a Flat turf summer effect) passed discovery: A/E 1.13 on 2045 bets (p 0.038), same-race control 0.97. O-b7f15d9f
   • H-cde96a9f (Rival: the trip-drop lift is a Flat turf summer effect) met the sealed vault supportive: A/E 1.05 on 8022 bets (p 0.0714, threshold 0.003). O-cf2f1651
   • H-eb945dd6 (Rival: stayers dropped to a sprint pay at non-stiff tracks too): the period 'after 2026-04-27' carries the effect (A/E 1.41) while the rest score 2025-10-01 to 2026-04-27 0.77. O-d5198642
2. AskedIs the after-April lift in trip-drop rules missing on Flat all-weather, which would confirm it is a turf-season effect rather than timing noise? answered
3. MechanismSummer turf ground and fields reward staying power in a way that winter all-weather racing does not. If the turf-season explanation is right, all-weather trip-droppers should run to their prices.
4. Why the market misses itIf the effect is turf-only, all-weather prices should already be fair. This test checks the turf explanation rather than looking for an edge.
5. Predictednegative, A/E 0.94 to 1.01 on 1500+ bets
6. Falsified ifA/E at or above 1.08 on all-weather, which would mean the lift is not tied to turf
7. Rule sketchrace_type = 'Flat' AND surface <> 'Turf' AND last_win_yards >= distance_yards + 330

The question's answer so far: Yes. On Flat all-weather the trip drop scored 0.96 on 1,058 bets (CI 0.77 to 1.18; GB 0.94), so the surface pattern fits a turf effect. But the turf rule itself fell to 1.05 in the vault, so there is little to explain.

Mechanism and rationale

Summer turf ground and fields reward staying power in a way that winter all-weather racing does not. If the turf-season explanation is right, all-weather trip-droppers should run to their prices.

The rule

race_type = 'Flat' AND surface <> 'Turf' AND last_win_yards >= distance_yards + 330

Reads: race_type, surface, last_win_yards, distance_yards

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

c1055d9bcc9649ec46d74b11b5d68f3bd238e0e69c21003c24e385cb000bb9c2

Frozen

not frozen

The definition can still change.

Confounders

• Surface and season overlap: all-weather racing is concentrated from October to March, while the turf lift came after April. A null on all-weather cannot separate 'turf' from 'summer'.
• All-weather has a different race mix: more low-class handicaps and more exposed horses. Class can move A/E on its own.
• Surface switchers: many all-weather trip-droppers won their longer race on turf, so the stamina was shown on a different surface.
• The vault already shrank the turf figure to 1.05, the same general drift seen at non-stiff tracks. There may be no turf effect for all-weather to lack.
• The lift in this family keeps landing at 16/1 or bigger. A few longshot wins could distort either result, so the price-capped refinement is needed.
• last_win_yards and last_win_distance_yards may not be the same measure (still unresolved). Previous rules used last_win_yards, so it is kept here for comparability.

Control

{
  "description": "Same-race control: the other runners in these all-weather races should come out at about 0.97–1.00 A/E. If the trip-droppers match that, the drop in trip carries no edge on all-weather. Contrast group: the same trip-drop rule on Flat turf. It scored 1.13 on 2,045 in discovery but shrank to 1.05 on 8,022 in the vault. The turf-season explanation needs all-weather to sit clearly below that turf figure, not just near 1.0.",
  "rule": "race_type = 'Flat' AND surface = 'Turf' AND last_win_yards >= distance_yards + 330"
}

Discovery result

fail broad rule · 16 Oct, 12:56

Bets

1,058

89 winners

A/E

0.96

95%: 0.77 to 1.18

SP return

-35.2%

level stakes

Betfair SP return

-19.5%

1,032 bets with a BSP

Strike rate 8.4%. p = 0.66. Same-race control: the other runners in these races had A/E 1.00, 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, 12:55

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND surface <> 'Turf' AND last_win_yards >= distance_yards + 330",
    "refinements": [
      {
        "label": "price-capped at 14/1 or shorter",
        "rule": "race_type = 'Flat' AND surface <> 'Turf' AND last_win_yards >= distance_yards + 330 AND sp_decimal <= 15.0"
      },
      {
        "label": "all-weather sprints only (mirrors the stiff-sprint system's trip band)",
        "rule": "race_type = 'Flat' AND surface <> 'Turf' AND last_win_yards >= distance_yards + 330 AND distance_yards <= 1340"
      }
    ],
    "control": {
      "description": "Same-race control: the other runners in these all-weather races should come out at about 0.97–1.00 A/E. If the trip-droppers match that, the drop in trip carries no edge on all-weather. Contrast group: the same trip-drop rule on Flat turf. It scored 1.13 on 2,045 in discovery but shrank to 1.05 on 8,022 in the vault. The turf-season explanation needs all-weather to sit clearly below that turf figure, not just near 1.0.",
      "rule": "race_type = 'Flat' AND surface = 'Turf' AND last_win_yards >= distance_yards + 330"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Surface and season overlap: all-weather racing is concentrated from October to March, while the turf lift came after April. A null on all-weather cannot separate 'turf' from 'summer'.",
      "All-weather has a different race mix: more low-class handicaps and more exposed horses. Class can move A/E on its own.",
      "Surface switchers: many all-weather trip-droppers won their longer race on turf, so the stamina was shown on a different surface.",
      "The vault already shrank the turf figure to 1.05, the same general drift seen at non-stiff tracks. There may be no turf effect for all-weather to lack.",
      "The lift in this family keeps landing at 16/1 or bigger. A few longshot wins could distort either result, so the price-capped refinement is needed.",
      "last_win_yards and last_win_distance_yards may not be the same measure (still unresolved). Previous rules used last_win_yards, so it is kept here for comparability."
    ],
    "mechanism": "Flat turf in summer runs on faster, more varied ground in bigger fields, and those races reward a horse that has proven stamina over further when it drops in trip. All-weather tracks are uniform and sharp, and winter all-weather racing favours speed and tactical position. Proven stamina should not give a horse the same advantage there. If the after-April trip-drop lift is a turf-season effect, all-weather trip-droppers should run only to their prices.",
    "market_miss": "This test checks a mechanism; it is not a search for an edge. If the effect belongs to turf, all-weather prices for trip-droppers should already be fair, because there is nothing for the market to miss. If all-weather shows the same lift, the turf/season explanation is wrong and the general trip-drop drift (about 1.05) is the better reading.",
    "prediction": {
      "direction": "negative",
      "size": "A/E about 0.94 to 1.01 on 1,500 or more bets, within about 0.03 of the same-race control and below the turf contrast",
      "falsified_if": "All-weather A/E of 1.08 or higher, or at or above the turf contrast group's A/E on the same window. Either would mean the lift is not tied to turf or the summer season."
    },
    "notes": "I checked the columns with query_catalog. surface ('turf or all-weather') and last_win_yards are pre-off. last_win_yards is available from 2020-09 in GB and IRE. The draft rule is kept unchanged so it can be compared directly with H-cde96a9f (the Flat turf rival). The direction is negative because the hypothesis predicts no lift. Under the lab's positive-direction pass criteria, the expected outcome is a non-pass. Read a pass as falsification of the turf explanation, not as an edge. The bet floor is raised to 300 so a null result is informative."
  }

Timeline

1. Job 1487 finished

   16 Oct, 12:56 · engine:compute-1

2. H-fdf540ac Discovery → KILLED fails: A/E 0.962, needs 0.95 or lower; p 0.3842 (lower tail), needs 0.05; periods [0.92, 1.12], each needs 0.99 or lower

   16 Oct, 12:56 · engine:evaluator

3. discovery result for H-fdf540ac (broad rule): fail

   16 Oct, 12:56 · engine:evaluator

4. H-fdf540ac Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 12:56 · engine:orchestrator

5. compute-1 took job 1487 (evaluate, attempt 1)

   16 Oct, 12:56 · compute-1

6. Job 1485 finished

   16 Oct, 12:56 · engine:ai-1

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

   16 Oct, 12:56 · engine:scheduler

8. H-fdf540ac Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 12:56 · agent:designer@v1

9. ai-1 took job 1485 (design, attempt 1)

   16 Oct, 12:55 · ai-1

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

    16 Oct, 12:55 · engine:scheduler

11. New idea H-fdf540ac: Test: the trip-drop lift is absent on Flat all-weather (test-mechanism)

    16 Oct, 12:55 · 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.