Small-field wide trips outside the 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-5ef5773d (Wide trip in a small field last time should pay less if lost ground is) passed discovery: A/E 1.39 on 349 bets (p 0.003), same-race control 0.92. O-5b199cb4
2. AskedIs the small-field edge carried by all-weather winter racing or by class, where small fields cluster? answered
3. MechanismIf small-field wide trips mark a tactical or muddling race whose bare form understates the horse, that should be true on turf as on the all-weather. If the edge is a winter all-weather pattern, it disappears on turf.
4. Why the market misses itSmall-field turf races are often slowly run, and a beaten margin under five lengths in them reads as true form. The 'wide' tag is in-running detail that the headline form line drops.
5. Predictedpositive, A/E 1.15 to 1.35 on 150+ bets
6. Falsified ifA/E 1.05 or lower on turf, which would make the small-field edge an all-weather effect
7. Rule sketchlast_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_field_size <= 8 AND lower(last_surface) = 'turf'
8. question idQ-406acee0
9. operatorreplicate

The question's answer so far: Partly. Outside the all-weather the small-field wide edge falls to 1.27 on 207 (p 0.07), and that rests on the period after April 2026 (1.47, against 0.92 before). So the all-weather bets and the timing carry much of the 1.39 small-field figure.

Mechanism and rationale

If small-field wide trips mark a tactical or muddling race whose bare form understates the horse, that should be true on turf as on the all-weather. If the edge is a winter all-weather pattern, it disappears on turf.

The rule

last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_field_size <= 8 AND lower(last_surface) = 'turf'

Reads: last_raced_wide, last_position, last_beaten_lengths, last_slow_start, last_field_size, last_surface

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

9b4fc980b7e757ce6b1b80cd319deb3981844b50de167710054d41137957cc24

Frozen

not frozen

The definition can still change.

Confounders

• Turf small fields over-represent better-class conditions/pattern races, where class rather than the wide trip may drive the result
• Jumps races are run on turf, so the broad rule mixes Hurdle/Chase small fields where 'wide' is often a chosen route on soft ground (that subset scored 0.97 on 41); the Flat-turf refinement isolates this
• Seasonal split: turf runs cluster in summer, and every wide-trip subset scored higher after April 2026, so check both periods
• Longshot artefact: wide-trip subsets carry ~2.0 A/E at 16/1+ on few wins; judge on A/E at 14/1 or shorter
• Selection from many splits of the wide-trip base: the small-field parent is one lead among many, so this replication inherits multiple-testing risk

Control

{
  "description": "Same-race other runners should score about 0.92-0.97 (the small-field parent's same-race control was 0.92), showing the edge is not a property of the races. Second comparison: the all-weather counterpart of the same small-field wide base; the hypothesis expects turf to score no worse than it. If turf is ~1.05 or below while the all-weather counterpart carries the 1.39 parent, the edge is an all-weather pattern.",
  "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_field_size <= 8 AND lower(last_surface) <> 'turf'"
}

Discovery result

fail broad rule · 16 Oct, 12:43

Bets

207

45 winners

A/E

1.27

95%: 0.92 to 1.69

SP return

+59.4%

level stakes

Betfair SP return

+145.2%

207 bets with a BSP

Strike rate 21.7%. p = 0.071. Same-race control: the other runners in these races had A/E 0.94, 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 · 36s · 16 Oct, 12:39

  Output (schema-checked)

  {
    "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_field_size <= 8 AND lower(last_surface) = 'turf'",
    "refinements": [
      {
        "label": "Flat turf only (like-for-like with all-weather)",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_field_size <= 8 AND lower(last_surface) = 'turf' AND last_race_type = 'Flat'"
      },
      {
        "label": "14/1 or shorter (strip the longshot band)",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_field_size <= 8 AND lower(last_surface) = 'turf' AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "Same-race other runners should score about 0.92-0.97 (the small-field parent's same-race control was 0.92), showing the edge is not a property of the races. Second comparison: the all-weather counterpart of the same small-field wide base; the hypothesis expects turf to score no worse than it. If turf is ~1.05 or below while the all-weather counterpart carries the 1.39 parent, the edge is an all-weather pattern.",
      "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_field_size <= 8 AND lower(last_surface) <> 'turf'"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.1,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1.05,
      "max_p": 0.2
    },
    "confounders": [
      "Turf small fields over-represent better-class conditions/pattern races, where class rather than the wide trip may drive the result",
      "Jumps races are run on turf, so the broad rule mixes Hurdle/Chase small fields where 'wide' is often a chosen route on soft ground (that subset scored 0.97 on 41); the Flat-turf refinement isolates this",
      "Seasonal split: turf runs cluster in summer, and every wide-trip subset scored higher after April 2026, so check both periods",
      "Longshot artefact: wide-trip subsets carry ~2.0 A/E at 16/1+ on few wins; judge on A/E at 14/1 or shorter",
      "Selection from many splits of the wide-trip base: the small-field parent is one lead among many, so this replication inherits multiple-testing risk"
    ],
    "mechanism": "A wide trip in a small field (8 or fewer runners) last time, beaten 5 lengths or less without a slow start, marks a tactical or muddling race whose bare form understates the horse. If that is the cause, it should hold on turf as on the all-weather; if the small-field edge is a winter all-weather pattern, it disappears when last time's race was on turf.",
    "market_miss": "Small-field turf races are often slowly run, and a beaten margin under five lengths in them reads as true form. The 'wide' tag is in-running detail that the headline form line drops, so the price is set off a form line that understates the horse.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.15 to 1.35 on 150+ bets; same-race control around 0.92-0.97",
      "falsified_if": "A/E 1.05 or lower on the turf rule (or on its 14/1-or-shorter version), especially if the all-weather counterpart keeps the parent's ~1.39. That would make the small-field edge an all-weather effect."
    },
    "notes": "I kept the rule sketch as written. last_surface is a pre-off column available from 2020-07-01 in GB and IRE. Turf includes jumps races, so the broad rule mixes in Hurdle and Chase small fields. The Flat-turf refinement is the like-for-like comparison with the all-weather, which only runs Flat. I tightened discovery to A/E 1.10, because the parent scored 1.39 and the prediction's lower edge is 1.15. Judge on A/E at 14/1 or shorter, not BSP ROI. The 16/1+ band has inflated every wide-trip subset. Report the before and after April 2026 periods separately."
  }

Timeline

1. Job 1405 finished

   16 Oct, 12:43 · engine:compute-1

2. H-3d0a7e42 Discovery → KILLED fails: p 0.0708, needs 0.05 or lower; periods [0.92, 1.47], each needs 1.0 (not a niche pass: a period points the other way [0.92, 1.47])

   16 Oct, 12:43 · engine:evaluator

3. discovery result for H-3d0a7e42 (broad rule): fail

   16 Oct, 12:43 · engine:evaluator

4. H-3d0a7e42 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 12:43 · engine:orchestrator

5. compute-1 took job 1405 (evaluate, attempt 1)

   16 Oct, 12:43 · compute-1

6. Job 1403 finished

   16 Oct, 12:39 · engine:ai-1

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

   16 Oct, 12:39 · engine:scheduler

8. H-3d0a7e42 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 12:39 · agent:designer@v1

9. ai-1 took job 1403 (design, attempt 1)

   16 Oct, 12:39 · ai-1

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

    16 Oct, 12:38 · engine:scheduler

11. New idea H-3d0a7e42: Small-field wide trips outside the all-weather (replicate)

    16 Oct, 12:38 · 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.