Wide trip last time on all-weather today still pays

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-463f02e9 (Wide trip at a galloping track last time should pay less if lost groun): the period 'after 2026-04-27' carries the effect (A/E 1.35) while the rest score 2025-10-01 to 2026-04-27 0.99. O-2f500cb7
   • H-c8ec6922 (Wide-trip edge outside the longshot band) passed discovery: A/E 1.21 on 974 bets (p 0.004), same-race control 0.95. O-487cb731
   • H-463f02e9 (Wide trip at a galloping track last time should pay less if lost groun) passed discovery as a niche: A/E 1.23 on 270 bets (p 0.069), same-race control 0.95. O-4f15022d
2. AskedIs the wide-trip edge present on all-weather racing today, or only on turf? answered
3. MechanismIf running wide undersells a horse's ability, the effect should carry to any surface. All-weather tracks are tight ovals where cover matters, so a horse that finished close up despite running wide should be underrated there too.
4. Why the market misses itThe form book shows a close-up placing with no excuse in the result. The 'raced wide' note sits in running comments that most bettors do not weigh, whatever the surface.
5. Predictedpositive, A/E 1.10 to 1.20 on 300+ bets
6. Falsified ifA/E 1.02 or lower on all-weather races today, which would mean the edge is a turf or summer effect and the live system needs a surface condition
7. Rule sketchlast_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND lower(surface) <> 'turf'

The question's answer so far: Yes, on both surfaces. On the same base, A/E was 1.23 on 462 bets on all-weather today and 1.25 on 554 on turf, against 1.00 on 8669 for close-up all-weather losers that were not wide.

Mechanism and rationale

If running wide undersells a horse's ability, the effect should carry to any surface. All-weather tracks are tight ovals where cover matters, so a horse that finished close up despite running wide should be underrated there too.

The rule

last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND lower(surface) <> 'turf'

Reads: last_raced_wide, last_position, last_beaten_lengths, last_slow_start, race_type, surface

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

8c11849c52f393cdf70cdd629ca8adaf5e0df13f304aa8242798271a9873c720

Frozen

not frozen

The definition can still change.

Confounders

• Calendar period: all-weather racing falls mostly between November 2025 and March 2026, so a null result could reflect the winter/early period (sibling subsets scored about 1.0 before April 2026) and not the surface itself. Check results split by month.
• Field size and race shape differ on all-weather (smaller fields, more front-running bias), which changes how often a wide trip happens and what it costs
• The parent's tilt exclusion (held-up or midfield runners from strongly tilted races) is dropped to match the sibling investigator rules (galloping, sharp, low-draw). A/E should be compared with those siblings and the 974-bet base, not with IQ140a's 1.27.
• Longshot artefact: horses beaten last time show a general lift at 16/1+, so the 14/1-or-shorter refinement is the cleaner read
• Surface switchers: horses that ran wide on turf last time and race on all-weather today carry a surface-change effect as well as the wide-trip trigger
• last_raced_wide exists only from November 2025, so there is about one winter of all-weather data and no independent third period

Discovery result

pass broad rule · 16 Oct, 12:20

Bets

462

92 winners

A/E

1.23

95%: 0.99 to 1.51

SP return

+0.3%

level stakes

Betfair SP return

+18.5%

460 bets with a BSP

Strike rate 19.9%. p = 0.029. Same-race control: the other runners in these races had A/E 0.95, 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: Beaten-distance cut tightened to 4 lengthspass

  378 bets · A/E 1.26 (1.00 to 1.56) · control 0.94 · 16 Oct, 12:22

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-272891.15

  after 2026-04-27891.53

• perturbation: Beaten-distance cut loosened to 6 lengthspass

  535 bets · A/E 1.29 (1.05 to 1.56) · control 0.94 · 16 Oct, 12:22

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 · 53s · 16 Oct, 12:18

  Output (schema-checked)

  {
    "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND lower(surface) <> 'turf'",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND lower(surface) <> 'turf' AND sp_decimal <= 15"
      },
      {
        "label": "wide on all-weather last time too",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND lower(surface) <> 'turf' AND lower(last_surface) <> 'turf'"
      }
    ],
    "control": {
      "description": "Same-race control: the other runners in the same all-weather races should show A/E about 0.95 to 1.00, so any lift belongs to the wide-trip selection and not to all-weather races in general. Surface comparison: the same base on turf today. The question is whether the edge needs turf, so the turf group is the comparison that matters. If turf shows about 1.2 and all-weather shows 1.02 or lower, the edge depends on surface or season. Because the all-weather races fall mostly before April 2026, the turf/all-weather split is also an early/late period split. Read it alongside the earlier finding that the period before April 2026 scored about 1.0 in sibling subsets.",
      "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND lower(surface) = 'turf'"
    },
    "pass_criteria": {
      "min_bets": 200,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Calendar period: all-weather racing falls mostly between November 2025 and March 2026, so a null result could reflect the winter/early period (sibling subsets scored about 1.0 before April 2026) and not the surface itself. Check results split by month.",
      "Field size and race shape differ on all-weather (smaller fields, more front-running bias), which changes how often a wide trip happens and what it costs",
      "The parent's tilt exclusion (held-up or midfield runners from strongly tilted races) is dropped to match the sibling investigator rules (galloping, sharp, low-draw). A/E should be compared with those siblings and the 974-bet base, not with IQ140a's 1.27.",
      "Longshot artefact: horses beaten last time show a general lift at 16/1+, so the 14/1-or-shorter refinement is the cleaner read",
      "Surface switchers: horses that ran wide on turf last time and race on all-weather today carry a surface-change effect as well as the wide-trip trigger",
      "last_raced_wide exists only from November 2025, so there is about one winter of all-weather data and no independent third period"
    ],
    "mechanism": "A horse that ran wide last time covered extra ground, so a close finish (beaten 5 lengths or less, not the winner, no slow start) undersells its ability. If the edge comes from ability being understated, and not from a turf-specific or summer pattern, it should carry to all-weather races today. All-weather tracks are tight ovals where cover counts, so a strong wide-trip run should be at least as informative there.",
    "market_miss": "The form figures and the result show only a close-up placing with no visible excuse. 'Raced wide' appears only in running comments, which most bettors and price-makers on low-grade all-weather cards do not weigh, so the price follows the bare beaten distance.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.10 to 1.20 on 300+ bets on all-weather races today",
      "falsified_if": "A/E 1.02 or lower on the broad all-weather rule, or a lift that comes only from 16/1+ runners (the 14/1-or-shorter refinement at 1.02 or lower). Either would mean the edge is a turf or summer effect and the live system needs a surface condition."
    },
    "notes": "I kept the rule sketch unchanged. The `surface` column ('turf' or 'all-weather') is pre-off and available from 2020, so the binding window is last_raced_wide from November 2025. NULL surface rows drop out under <>. I raised min_bets to 200 because the prediction is 300+ bets and the test is short. The turf-today control rule is the main comparison for the question. The refinements keep the full broad base, so only one condition changes in each."
  }

Timeline

1. Job 1317 finished

   16 Oct, 12:22 · engine:compute-1

2. H-5b9fa0d3 Falsification → KILLED falsified: IRE A/E 0.9

   16 Oct, 12:22 · engine:falsification

3. falsification result for H-5b9fa0d3 (partition: Handicaps only): pass

   16 Oct, 12:22 · engine:evaluator

4. compute-1 took job 1317 (attack, attempt 1)

   16 Oct, 12:22 · compute-1

5. Job 1316 finished

   16 Oct, 12:22 · engine:compute-1

6. falsification result for H-5b9fa0d3 (partition: No surface switch: all-weather last time and today): pass

   16 Oct, 12:22 · engine:evaluator

7. compute-1 took job 1316 (attack, attempt 1)

   16 Oct, 12:22 · compute-1

8. Job 1315 finished

   16 Oct, 12:22 · engine:compute-1

9. falsification result for H-5b9fa0d3 (price_band: Middle price band removed: 7/1 or shorter only): pass

   16 Oct, 12:22 · engine:evaluator

10. compute-1 took job 1315 (attack, attempt 1)

    16 Oct, 12:22 · compute-1

11. Job 1314 finished

    16 Oct, 12:22 · engine:compute-1

12. falsification result for H-5b9fa0d3 (control: Control: close-up last time on all-weather but NOT wide): pass

    16 Oct, 12:22 · engine:evaluator

13. compute-1 took job 1314 (attack, attempt 1)

    16 Oct, 12:22 · compute-1

14. Job 1313 finished

    16 Oct, 12:22 · engine:compute-1

15. falsification result for H-5b9fa0d3 (perturbation: Beaten-distance cut loosened to 6 lengths): pass

    16 Oct, 12:22 · engine:evaluator

16. compute-1 took job 1313 (attack, attempt 1)

    16 Oct, 12:22 · compute-1

17. Job 1312 finished

    16 Oct, 12:22 · engine:compute-1

18. falsification result for H-5b9fa0d3 (perturbation: Beaten-distance cut tightened to 4 lengths): pass

    16 Oct, 12:22 · engine:evaluator

19. compute-1 took job 1312 (attack, attempt 1)

    16 Oct, 12:22 · compute-1

20. Job 1305 finished

    16 Oct, 12:21 · engine:ai-1

21. Queued attack on the compute lane: attack: Handicaps only

    16 Oct, 12:21 · engine:scheduler