Rival: the big-field lift is a drop into a smaller race today, not a misleading margin

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-d97a030e (Rival: big-field beatings, not the heavy beating itself, carry the lif) survived falsification: 6 of 6 attacks left it standing; it is frozen for its one vault query. O-09d8e026
   • H-d97a030e (Rival: big-field beatings, not the heavy beating itself, carry the lif) passed discovery: A/E 1.19 on 2097 bets (p 0.022), same-race control 0.98. O-93b1b6e2
2. AskedIs the big-field heavy-beating lift a step down in competition (a smaller, weaker race today) rather than beaten lengths overstated by the size of the field? answered
3. MechanismHorses beaten in big, competitive festival-type handicaps often race next in ordinary, smaller handicaps. There they meet weaker rivals, so the lift could come from the easier race and not from the margin misleading anyone.
4. Why the market misses itIf the rival is right, the market under-weights the drop in race strength because the horse's last figure looks so bad. If it is wrong, the lift should be the same whatever today's field size.
5. Predictedpositive, A/E 1.25 or more on 500+ bets if the rival holds, clearly above big-field beaten runners now in fields of 14 or more
6. Falsified ifA/E at or below the big-field parent's 1.187, or no higher than the same group running in fields of 14 or more today. That would rule the rival out and keep the misleading-margin mechanism.
7. Rule sketchis_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size <= 10

The question's answer so far: No. The smaller-field-today cut scored 1.15 on 526, below the parent's 1.187, and was killed for resting on the up-to-7/1 band alone (rest 0.88). A step down in field size does not carry the big-field lift.

Mechanism and rationale

Horses beaten in big, competitive festival-type handicaps often race next in ordinary, smaller handicaps. There they meet weaker rivals, so the lift could come from the easier race and not from the margin misleading anyone.

The rule

is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size <= 10

Reads: is_handicap, race_type, last_beaten_lengths, last_field_size, field_size

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

cc95329686a5715ea13d16cd7507c3faae0e53f676df547b97c85394238e1eb7

Frozen

not frozen

The definition can still change.

Confounders

• Region: most big-field heavy-beating bets are Irish, so the result should be reported by country and must not rest on one
• Price band: smaller fields shorten prices and move runners out of the 16/1+ band, where the parent's own lift behaves differently. Compare like-for-like with the under-16/1 refinement
• Small-field races may be over-round or under-round in different ways. Check book_pct and the same-race control
• Last runs cluster at festivals, so this year's festival form and later seasonality are confounded
• Going change between runs (e.g. festival softness, then better ground)
• Trainer placement skill: yards that aim at a big festival handicap and then place the horse in a weak race
• Days since the last run: a quick return after a festival differs from a long break

Control

{
  "description": "The deciding comparison is the same heavily beaten big-field group running today in fields of 14 or more. If the rival holds, the rule must clearly beat this control. If both sit near the parent's 1.187, the lift does not depend on today's field size, which rules the rival out. The lab's standard same-race control (the other runners in the rule's races) should come out near 1.0, at about 0.99 like the parent's control. A same-race control well above 1 would point to a property of small handicaps in general rather than of the horses.",
  "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size >= 14"
}

Discovery result

pass broad rule (niche) · 17 Oct, 03:29

Bets

526

60 winners

A/E

1.15

95%: 0.88 to 1.48

SP return

-13.7%

level stakes

Betfair SP return

+11.1%

526 bets with a BSP

Strike rate 11.4%. p = 0.15. Same-race control: the other runners in these races had A/E 0.99, 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.

• control: Deciding control: same group in big fields today (14+)pass

  1,136 bets · A/E 1.24 (0.95 to 1.60) · control 0.97 · 17 Oct, 03:30

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-275241.07

  after 2026-04-276121.38

• control: Control without the big last field: heavy beating in a small race, now in a small racepass

  1,870 bets · A/E 1.07 (0.92 to 1.23) · control 1.00 · 17 Oct, 03:30

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 · 32s · 17 Oct, 03:26

  Output (schema-checked)

  {
    "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size <= 10",
    "refinements": [
      {
        "label": "Under 16/1 today (the band where the big-field lift sits)",
        "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size <= 10 AND sp_decimal < 17"
      },
      {
        "label": "Sharpest step down in field: 8 or fewer runners today",
        "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size <= 8"
      },
      {
        "label": "Handicap to handicap only (like-for-like drop in race strength)",
        "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size <= 10 AND last_is_handicap = 1"
      }
    ],
    "control": {
      "description": "The deciding comparison is the same heavily beaten big-field group running today in fields of 14 or more. If the rival holds, the rule must clearly beat this control. If both sit near the parent's 1.187, the lift does not depend on today's field size, which rules the rival out. The lab's standard same-race control (the other runners in the rule's races) should come out near 1.0, at about 0.99 like the parent's control. A same-race control well above 1 would point to a property of small handicaps in general rather than of the horses.",
      "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND field_size >= 14"
    },
    "pass_criteria": {
      "min_bets": 200,
      "min_ae": 1.1,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Region: most big-field heavy-beating bets are Irish, so the result should be reported by country and must not rest on one",
      "Price band: smaller fields shorten prices and move runners out of the 16/1+ band, where the parent's own lift behaves differently. Compare like-for-like with the under-16/1 refinement",
      "Small-field races may be over-round or under-round in different ways. Check book_pct and the same-race control",
      "Last runs cluster at festivals, so this year's festival form and later seasonality are confounded",
      "Going change between runs (e.g. festival softness, then better ground)",
      "Trainer placement skill: yards that aim at a big festival handicap and then place the horse in a weak race",
      "Days since the last run: a quick return after a festival differs from a long break"
    ],
    "mechanism": "Horses beaten 20 lengths or more in big (14+ runner), competitive, often festival-type jumps handicaps often turn out next in ordinary, smaller handicaps of 10 or fewer runners. There the rivals are weaker and there is less trouble in running. On this rival, the parent's big-field lift comes from the step down in competition today, not from a margin that overstates how badly the horse ran.",
    "market_miss": "Bettors anchor on the ugly beaten distance and under-weight how much weaker today's small race is than the deep handicap the horse was beaten in. If the misleading-margin story is right instead, the lift should be the same whatever today's field size, and this rule should not beat its 14+ field control.",
    "prediction": {
      "direction": "positive",
      "size": "A/E about 1.25 or more on several hundred bets, clearly above the same big-field group running in fields of 14 or more today",
      "falsified_if": "A/E at or below the big-field parent's 1.187, or no higher than the control (the same group in fields of 14 or more today). Either would rule the rival out and keep the misleading-margin mechanism."
    },
    "notes": "I kept the rule sketch as it was: all its columns are pre-off and it slices only on today's field size against the parent. Pass criteria are stricter than the lab floor (200 bets, A/E 1.10), because a rival must clearly beat the parent's 1.187 to mean anything. Even if the absolute test passes, the rival is supported only if the rule also beats the 14+ field control. Report both countries separately."
  }

Timeline

1. Job 2692 finished

   17 Oct, 03:30 · engine:compute-1

2. H-91a89a7c Falsification → KILLED falsified: driven by one price band (up to 7/1): the rest score 0.88

   17 Oct, 03:30 · engine:falsification

3. falsification result for H-91a89a7c (partition: Partition: chases only): pass

   17 Oct, 03:30 · engine:evaluator

4. compute-1 took job 2692 (attack, attempt 1)

   17 Oct, 03:30 · compute-1

5. Job 2691 finished

   17 Oct, 03:30 · engine:compute-1

6. falsification result for H-91a89a7c (perturbation: Perturbation loosened: beaten 16+, last field 12+, today 11 or fewer): pass

   17 Oct, 03:30 · engine:evaluator

7. compute-1 took job 2691 (attack, attempt 1)

   17 Oct, 03:30 · compute-1

8. Job 2690 finished

   17 Oct, 03:30 · engine:compute-1

9. falsification result for H-91a89a7c (perturbation: Perturbation tightened: beaten 24+, last field 16+, today 8 or fewer): pass

   17 Oct, 03:30 · engine:evaluator

10. compute-1 took job 2690 (attack, attempt 1)

    17 Oct, 03:30 · compute-1

11. Job 2689 finished

    17 Oct, 03:30 · engine:compute-1

12. falsification result for H-91a89a7c (control: Dose-response: mid-size field today (11-13) should sit between rule and big-field control): pass

    17 Oct, 03:30 · engine:evaluator

13. compute-1 took job 2689 (attack, attempt 1)

    17 Oct, 03:30 · compute-1

14. Job 2688 finished

    17 Oct, 03:30 · engine:compute-1

15. falsification result for H-91a89a7c (control: Control without the big last field: heavy beating in a small race, now in a small race): pass

    17 Oct, 03:30 · engine:evaluator

16. compute-1 took job 2688 (attack, attempt 1)

    17 Oct, 03:30 · compute-1

17. Job 2687 finished

    17 Oct, 03:30 · engine:compute-1

18. falsification result for H-91a89a7c (control: Deciding control: same group in big fields today (14+)): pass

    17 Oct, 03:30 · engine:evaluator

19. compute-1 took job 2687 (attack, attempt 1)

    17 Oct, 03:30 · compute-1