Test-mechanism: big-field jumps beatings where the horse was eased carry the lift

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-7358e6fb (Replicate: Flat handicappers heavily beaten in a big field last time b) was predicted positive but came out the other way: A/E 0.95 on 1560 bets (p 0.29 for the opposite direction). O-0cca5f69
2. AskedIs the big-field heavy-beating lift jumps-only because jumps margins are inflated by easing and tiredness once beaten, which Flat races largely lack? testing
3. MechanismIn big-field jumps handicaps, a beaten horse is often eased right down over the last half-mile. The recorded margin then piles up lengths that say nothing about its ability. Flat jockeys ease less, and Flat trips are shorter, which fits the Flat replicate's failure.
4. Why the market misses itThe form figures and the beaten distance show a 25-length hammering. The fact that it was eased sits in the comment, and the price follows the margin.
5. Predictedpositive, A/E 1.20 to 1.35 on 400+ bets, above the big-field parent's 1.187
6. Falsified ifA/E at or below 1.187, or no higher than the same rule with last_eased = 0
7. Rule sketchis_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_eased = 1
8. question idQ-d422b625
9. operator

Mechanism and rationale

In big-field jumps handicaps, a beaten horse is often eased right down over the last half-mile. The recorded margin then piles up lengths that say nothing about its ability. Flat jockeys ease less, and Flat trips are shorter, which fits the Flat replicate's failure.

The rule

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

Reads: is_handicap, race_type, last_beaten_lengths, last_field_size, last_eased

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

9f80d834f515c14afa41543dcc9f823aeb5c3770c1b30876a3e65cacc846c77c

Frozen

not frozen

The definition can still change.

Confounders

• Eased horses may have had a physical problem (bled, went lame, broke a blood vessel) and come back unsound or off a long break. Check by days_since_last_run.
• last_eased depends on comment wording, which may be written differently in GB and IRE. Report each country separately.
• Being eased overlaps with near pull-ups, long trips and testing going. The lift may belong to trip or going, not to the inflated margin.
• Eased horses go off at bigger prices, and the family's lift dies at 16/1 or bigger. The price mix can hide or fake the effect, so read it next to the under-16/1 refinement.
• last_eased and last_beaten_lengths only start in November 2025, so the window is short and one season could drive the result.
• A class drop or a mark drop after the beating could carry the lift, not the easing itself.

Control

{
  "description": "The main control is the same big-field heavy-beating rule with last_eased = 0. The mechanism says easing is what inflates the margin, so the eased group should beat the non-eased group clearly, and the non-eased group should sit at or below the parent's 1.187. If both groups score about the same, easing does not explain the lift. Other runners in the same races should come out near 1.0, which shows today's race is not just generous to outsiders.",
  "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_eased = 0"
}

Discovery result

fail broad rule · 17 Oct, 03:42

Bets

18

2 winners

A/E

1.76

95%: 0.21 to 6.35

SP return

+111.1%

level stakes

Betfair SP return

–

18 bets with a BSP

Strike rate 11.1%. p = 0.31. Same-race control: the other runners in these races had A/E 0.96, 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 · 47s · 17 Oct, 03:37

  Output (schema-checked)

  {
    "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_eased = 1",
    "refinements": [
      {
        "label": "Under 16/1 (where the big-field lift lives)",
        "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_eased = 1 AND sp_decimal < 17"
      },
      {
        "label": "Easing happened in a jumps race last time",
        "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_eased = 1 AND last_race_type IN ('Hurdle', 'Chase')"
      }
    ],
    "control": {
      "description": "The main control is the same big-field heavy-beating rule with last_eased = 0. The mechanism says easing is what inflates the margin, so the eased group should beat the non-eased group clearly, and the non-eased group should sit at or below the parent's 1.187. If both groups score about the same, easing does not explain the lift. Other runners in the same races should come out near 1.0, which shows today's race is not just generous to outsiders.",
      "rule": "is_handicap = 1 AND race_type IN ('Hurdle', 'Chase') AND last_beaten_lengths >= 20 AND last_field_size >= 14 AND last_eased = 0"
    },
    "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": [
      "Eased horses may have had a physical problem (bled, went lame, broke a blood vessel) and come back unsound or off a long break. Check by days_since_last_run.",
      "last_eased depends on comment wording, which may be written differently in GB and IRE. Report each country separately.",
      "Being eased overlaps with near pull-ups, long trips and testing going. The lift may belong to trip or going, not to the inflated margin.",
      "Eased horses go off at bigger prices, and the family's lift dies at 16/1 or bigger. The price mix can hide or fake the effect, so read it next to the under-16/1 refinement.",
      "last_eased and last_beaten_lengths only start in November 2025, so the window is short and one season could drive the result.",
      "A class drop or a mark drop after the beating could carry the lift, not the easing itself."
    ],
    "mechanism": "In big-field jumps handicaps, a horse that is beaten is often eased right down over the last half-mile. Each length lost after that is a choice, not a sign of ability, so the recorded margin (20+ lengths) overstates how inferior it is. Flat jockeys ease less, and over shorter trips there is less ground to lose, which fits the failed Flat replicate. If this is the mechanism, the big-field jumps lift should sit mostly in the eased runs.",
    "market_miss": "Form figures, beaten-distance summaries and ratings-based tools take the 20-plus-length margin at face value. 'Eased' appears only in the race comment, so the price follows the margin and does not discount the lengths given away after the horse was eased.",
    "prediction": {
      "direction": "positive",
      "size": "A/E about 1.20 to 1.35, above the big-field parent's 1.187, and likely on fewer than the 400 bets the hypothesis hoped for, given the short window",
      "falsified_if": "Falsified if A/E is at or below 1.187, or not clearly above the last_eased = 0 control (the same rule otherwise). Also falsified if the lift appears in only one country, or only at 16/1 or bigger."
    },
    "notes": "I kept the rule sketch unchanged. The catalogue says last_eased and last_beaten_lengths only start on 2025-11-01 in both GB and IRE, so the window is short and the sample may be small. A thin result that points the right way should go to falsification and not be read as a kill. The key comparison is the eased versus not-eased split; beating the parent alone is not enough. I left out the trip refinement because the reasoning does not set a cut-off."
  }

Timeline

1. Job 2721 finished

   17 Oct, 03:42 · engine:compute-1

2. H-52702d58 Discovery → KILLED fails: 18 bets, needs 150; p 0.3145, needs 0.05 or lower; periods [], each needs 1.0 (not a niche pass: only 18 bets)

   17 Oct, 03:42 · engine:evaluator

3. discovery result for H-52702d58 (broad rule): fail

   17 Oct, 03:42 · engine:evaluator

4. H-52702d58 Designed → RESEARCHdiscovery measuring on the discovery zone

   17 Oct, 03:42 · engine:orchestrator

5. compute-1 took job 2721 (evaluate, attempt 1)

   17 Oct, 03:42 · compute-1

6. Job 2719 finished

   17 Oct, 03:38 · engine:ai-1

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

   17 Oct, 03:38 · engine:scheduler

8. H-52702d58 Proposed → RESEARCHdesigned pre-spec written and hashed

   17 Oct, 03:38 · agent:designer@v1

9. ai-1 took job 2719 (design, attempt 1)

   17 Oct, 03:37 · ai-1

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

    17 Oct, 03:37 · engine:scheduler

11. New idea H-52702d58: Test-mechanism: big-field jumps beatings where the horse was eased carry the lift (test-mechanism)

    17 Oct, 03:37 · 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.