Wide-trip base in jump races today: is the edge only a summer Flat turf effect?

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-3d0a7e42 (Small-field wide trips outside the all-weather) pointed the predicted way but failed the bar: A/E 1.27 on 207 bets (p 0.071), control 0.94. O-194cb689
   • H-3d0a7e42 (Small-field wide trips outside the all-weather): the period 'after 2026-04-27' carries the effect (A/E 1.47) while the rest score 2025-10-01 to 2026-04-27 0.92. O-b4f84fea
2. AskedIs the lift after April 2026, seen in several wide-trip subsets, a summer turf effect, or a change in how often and where the 'wide' tag gets recorded? testing
3. MechanismIf the lift is seasonal, or follows a change in Flat turf tagging, it should be missing in jump races, which run mostly in winter. If wide runners still pay over jumps, the edge is not seasonal.
4. Why the market misses itIn jump races a wide, beaten run reads as a poor effort, just as on the Flat, so the market should misprice it there too if the effect is real.
5. Predictedpositive, A/E 1.10 to 1.20 on 250+ bets at 15.0 SP or shorter
6. Falsified ifA/E 1.02 or lower over jumps today. The edge would then belong to the Flat season.
7. Rule sketchlast_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15
8. question id

Mechanism and rationale

If the lift is seasonal, or follows a change in Flat turf tagging, it should be missing in jump races, which run mostly in winter. If wide runners still pay over jumps, the edge is not seasonal.

The rule

last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15

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

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

680b0e47f20226ba632bb7901d0d9e8796c6f1c951f4438c39f048dd963e6fad

Frozen

not frozen

The definition can still change.

Confounders

• Deliberately wide routes on soft or heavy ground over jumps (H-fc98577b scored 0.97), which would dilute a real effect; refinement 2 excludes them
• Comment conventions differ over jumps: 'raced wide' may be tagged less often or for other reasons than on the Flat, so the tag may not mean the same thing
• Many of today's jump runners had a Flat or bumper last run, which mixes tagging codes; refinement 1 isolates the jump-to-jump case
• Festival concentration: big-meeting fields and prices could dominate a thin sample
• Jump racing is mostly in winter, so a seasonal effect and a code effect cannot be fully separated; summer jumps are a small subset
• Thin sample: the jump subset may fall short of 250 bets, so a near-1.0 reading would be underpowered

Control

{
  "description": "Other runners in the same jump races should score about 0.92-1.00 A/E; the selection must clear them clearly, or the result reflects the race and not the wide tag. Comparison group: the same wide-trip base at 15.0 SP or shorter on the Flat today (the source of the post-April-2026 lift). It shows whether the jumps A/E is in line with the Flat base (about 1.2) or falls short of it. Expected: the Flat control is around 1.2, with its lift concentrated after April 2026.",
  "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND sp_decimal <= 15"
}

Discovery result

pass broad rule (niche) · 16 Oct, 12:56

Bets

130

31 winners

A/E

1.27

95%: 0.86 to 1.80

SP return

+8.3%

level stakes

Betfair SP return

+30.0%

130 bets with a BSP

Strike rate 23.8%. p = 0.11. 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

Each attack is a deliberate attempt to kill the idea. 5 of 6 survived.

• perturbation: Tighter cuts: beaten 4L or less, SP 12.0 or shorterpass

  101 bets · A/E 1.37 (0.91 to 1.98) · control 0.91 · 16 Oct, 12:58

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27661.38

  after 2026-04-2735–

• perturbation: Looser cuts: beaten 6L or less, SP 18.0 or shorterpass

  160 bets · A/E 1.24 (0.87 to 1.73) · control 0.95 · 16 Oct, 12:58

  Breakdown

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, 12:53

  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 IN ('Hurdle', 'Chase') AND sp_decimal <= 15",
    "refinements": [
      {
        "label": "like-for-like: wide tag earned over jumps",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15 AND last_race_type IN ('Hurdle', 'Chase')"
      },
      {
        "label": "exclude deliberate soft-ground wide routes (H-fc98577b confounder)",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group NOT IN ('soft', 'heavy')"
      }
    ],
    "control": {
      "description": "Other runners in the same jump races should score about 0.92-1.00 A/E; the selection must clear them clearly, or the result reflects the race and not the wide tag. Comparison group: the same wide-trip base at 15.0 SP or shorter on the Flat today (the source of the post-April-2026 lift). It shows whether the jumps A/E is in line with the Flat base (about 1.2) or falls short of it. Expected: the Flat control is around 1.2, with its lift concentrated after April 2026.",
      "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND race_type = 'Flat' AND sp_decimal <= 15"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Deliberately wide routes on soft or heavy ground over jumps (H-fc98577b scored 0.97), which would dilute a real effect; refinement 2 excludes them",
      "Comment conventions differ over jumps: 'raced wide' may be tagged less often or for other reasons than on the Flat, so the tag may not mean the same thing",
      "Many of today's jump runners had a Flat or bumper last run, which mixes tagging codes; refinement 1 isolates the jump-to-jump case",
      "Festival concentration: big-meeting fields and prices could dominate a thin sample",
      "Jump racing is mostly in winter, so a seasonal effect and a code effect cannot be fully separated; summer jumps are a small subset",
      "Thin sample: the jump subset may fall short of 250 bets, so a near-1.0 reading would be underpowered"
    ],
    "mechanism": "A recent wide trip, beaten 5 lengths or less without a slow start, understates the horse's ability, so it goes off overpriced next time. If the existing edge (1.21 on 974 at 15.0 SP or shorter) were seasonal, or came from a change in how Flat turf comments are tagged, it would be missing in today's jump races, which run mostly in winter. If wide runners still beat their prices over jumps, the edge is neither seasonal nor specific to Flat tagging.",
    "market_miss": "Over jumps the market reads a narrowly beaten run with a wide trip as an ordinary losing effort, as it does on the Flat. It does not credit the ground or position the trip cost, so if the effect is a real feature of the horse rather than a Flat-season artefact, the price will be too long in jump races too.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.10 to 1.20 on 250+ bets at 15.0 SP or shorter in hurdles and chases",
      "falsified_if": "A/E 1.02 or lower over jumps today on an adequate sample (150+ bets), or same-race other runners scoring as well as the selection. The edge would then belong to the Flat season or Flat tagging."
    },
    "notes": "The draft rule is kept unchanged: all its columns are pre-off and the price cap matches the 15.0 SP base finding. No calendar-month filter is used, because race_type already serves as the winter proxy the question needs. The Flat control rule is for comparison only. If the jumps sample is thin but points the right way, it should go on to falsification rather than be killed."
  }

Timeline

1. Job 1502 finished

   16 Oct, 12:58 · engine:compute-1

2. H-77a1708f Falsification → KILLED falsified: IRE A/E 0.913

   16 Oct, 12:58 · engine:falsification

3. falsification result for H-77a1708f (partition: Jump handicaps only): pass

   16 Oct, 12:58 · engine:evaluator

4. compute-1 took job 1502 (attack, attempt 1)

   16 Oct, 12:58 · compute-1

5. Job 1501 finished

   16 Oct, 12:58 · engine:compute-1

6. falsification result for H-77a1708f (control: Control: same close-beaten jump runners without a wide trip): fail

   16 Oct, 12:58 · engine:evaluator

7. compute-1 took job 1501 (attack, attempt 1)

   16 Oct, 12:58 · compute-1

8. Job 1500 finished

   16 Oct, 12:58 · engine:compute-1

9. falsification result for H-77a1708f (partition: Like-for-like: wide tag earned in a jump race last time): pass

   16 Oct, 12:58 · engine:evaluator

10. compute-1 took job 1500 (attack, attempt 1)

    16 Oct, 12:58 · compute-1

11. Job 1499 finished

    16 Oct, 12:58 · engine:compute-1

12. falsification result for H-77a1708f (price_band: Shorter prices only (8.0 SP or less), without the 14/1 band): pass

    16 Oct, 12:58 · engine:evaluator

13. compute-1 took job 1499 (attack, attempt 1)

    16 Oct, 12:58 · compute-1

14. Job 1498 finished

    16 Oct, 12:58 · engine:compute-1

15. falsification result for H-77a1708f (perturbation: Looser cuts: beaten 6L or less, SP 18.0 or shorter): pass

    16 Oct, 12:58 · engine:evaluator

16. compute-1 took job 1498 (attack, attempt 1)

    16 Oct, 12:58 · compute-1

17. Job 1497 finished

    16 Oct, 12:58 · engine:compute-1

18. falsification result for H-77a1708f (perturbation: Tighter cuts: beaten 4L or less, SP 12.0 or shorter): pass

    16 Oct, 12:58 · engine:evaluator

19. compute-1 took job 1497 (attack, attempt 1)

    16 Oct, 12:58 · compute-1

20. Job 1489 finished

    16 Oct, 12:57 · engine:ai-1

21. Queued attack on the compute lane: attack: Jump handicaps only

    16 Oct, 12:57 · engine:scheduler