Generalise: big-field heavy beatings mislead in non-handicap jumps races too

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) passed discovery: A/E 1.19 on 2097 bets (p 0.022), same-race control 0.98. O-93b1b6e2
2. AskedDoes a heavy beating in a big field mislead the market outside handicaps too, for example novice and maiden hurdlers? testing
3. MechanismEasing and traffic in a 14+ runner field inflate the margin whatever today's race conditions, so the effect should not depend on the next race being a handicap.
4. Why the market misses itNovice and maiden hurdlers are priced heavily off their visible form figures and margins, with no handicap mark to anchor them, so an inflated margin should hurt their price at least as much.
5. Predictedpositive, A/E 1.10 to 1.25 on 600+ bets, in both GB and IRE
6. Falsified ifA/E below 1.05, or IRE below 0.95, or no higher than non-handicap runners beaten 20+ lengths in last fields of 10 or fewer
7. Rule sketchis_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14
8. question idQ-bcf47e28
9. operatorgeneralise

Mechanism and rationale

Easing and traffic in a 14+ runner field inflate the margin whatever today's race conditions, so the effect should not depend on the next race being a handicap.

The rule

is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14

Reads: is_handicap, race_type, last_beaten_lengths, last_field_size

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

135de304383149a68e1a25e00f30c40a09091652dae549bf97e3f41119c234c3

Frozen

not frozen

The definition can still change.

Confounders

• Big-yard novices at short prices dominate non-handicaps; the lift could be a trainer/yard effect, not a margin effect
• Bumper- or flat-to-hurdle switchers: a heavy beating in a big bumper says little about jumping ability (refinement 2 isolates this)
• Longshot band behaviour: the family's 16/1+ lifts look like a separate effect from the competitive-price big-field lift (refinement 1)
• Region mix: earlier narrow cuts produced GB-only effects; IRE must hold on its own
• Seasonality of novice campaigns (autumn debut runs, spring festivals)
• Field size today correlates with last field size; big-field races today carry their own price structure

Control

{
  "description": "Same-race other runners should be near par (about 0.97-1.00). The question's falsification test also needs a small-field comparison: non-handicap jumps runners beaten 20+ lengths last time in a field of 10 or fewer. If the mechanism holds (easing and traffic in big fields inflate the margin), this group should be clearly lower, around 1.00 or below, and at least 0.05 A/E under the main rule.",
  "rule": "is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size <= 10"
}

Designer

Discovery result

fail broad rule · 17 Oct, 03:30

Bets

1,684

83 winners

A/E

1.06

95%: 0.84 to 1.31

SP return

-62.7%

level stakes

Betfair SP return

-45.8%

1,684 bets with a BSP

Strike rate 4.9%. p = 0.32. Same-race control: the other runners in these races had A/E 0.97, 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 · 26s · 17 Oct, 03:27

  Output (schema-checked)

  {
    "rule": "is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND sp_decimal <= 15"
      },
      {
        "label": "last run over jumps (excludes bumper/flat-to-hurdle switchers)",
        "rule": "is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_race_type IN ('Hurdle', 'Chase')"
      }
    ],
    "control": {
      "description": "Same-race other runners should be near par (about 0.97-1.00). The question's falsification test also needs a small-field comparison: non-handicap jumps runners beaten 20+ lengths last time in a field of 10 or fewer. If the mechanism holds (easing and traffic in big fields inflate the margin), this group should be clearly lower, around 1.00 or below, and at least 0.05 A/E under the main rule.",
      "rule": "is_handicap = 0 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size <= 10"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 150,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Big-yard novices at short prices dominate non-handicaps; the lift could be a trainer/yard effect, not a margin effect",
      "Bumper- or flat-to-hurdle switchers: a heavy beating in a big bumper says little about jumping ability (refinement 2 isolates this)",
      "Longshot band behaviour: the family's 16/1+ lifts look like a separate effect from the competitive-price big-field lift (refinement 1)",
      "Region mix: earlier narrow cuts produced GB-only effects; IRE must hold on its own",
      "Seasonality of novice campaigns (autumn debut runs, spring festivals)",
      "Field size today correlates with last field size; big-field races today carry their own price structure"
    ],
    "mechanism": "In a field of 14 or more, a beaten horse is often eased once its chance has gone, or loses ground in traffic, so its beaten margin overstates how bad the run was. This inflation comes from the last race, not today's, so it should carry into non-handicap hurdles and chases (novice, maiden, conditions) just as it did in handicaps (1.187 on 2097 for 14+ runners in the handicap rule).",
    "market_miss": "Non-handicap jumpers have no official mark to anchor them, so they are priced heavily off visible form figures and margins. A 20+ length beating reads as plain inability, and the market does not discount it for the size of the field. So the price should be pushed out at least as much as in handicaps.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.10 to 1.25 on 600+ bets, with GB and IRE both above par",
      "falsified_if": "Falsified if overall A/E is below 1.05, or IRE A/E is below 0.95, or A/E is no higher than the small-field control (non-handicap jumps runners beaten 20+ lengths in last fields of 10 or fewer)."
    },
    "notes": "I kept the rule sketch as it was: all its columns are pre-off, and it does not read the result. Requiring last_beaten_lengths >= 20 means last-time non-finishers (NULL margin) are left out. I raised min_bets to 300 because the hypothesis predicts 600+ bets. The family findings put the big-field lift under 16/1, so the price refinement uses sp_decimal <= 15."
  }

Timeline

1. Job 2672 finished

   17 Oct, 03:30 · engine:compute-1

2. H-f316faf4 Discovery → KILLED fails: p 0.3223, needs 0.05 or lower; periods [0.94, 1.16], each needs 1.0 (not a niche pass: a period points the other way [0.94, 1.16])

   17 Oct, 03:30 · engine:evaluator

3. discovery result for H-f316faf4 (broad rule): fail

   17 Oct, 03:30 · engine:evaluator

4. H-f316faf4 Designed → RESEARCHdiscovery measuring on the discovery zone

   17 Oct, 03:30 · engine:orchestrator

5. compute-1 took job 2672 (evaluate, attempt 1)

   17 Oct, 03:30 · compute-1

6. Job 2669 finished

   17 Oct, 03:28 · engine:ai-1

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

   17 Oct, 03:28 · engine:scheduler

8. H-f316faf4 Proposed → RESEARCHdesigned pre-spec written and hashed

   17 Oct, 03:28 · agent:designer@v1

9. ai-1 took job 2669 (design, attempt 1)

   17 Oct, 03:27 · ai-1

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

    17 Oct, 03:26 · engine:scheduler

11. New idea H-f316faf4: Generalise: big-field heavy beatings mislead in non-handicap jumps races too (generalise)

    17 Oct, 03:26 · 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.