Rival: poor recent form at 14/1 or shorter beats its price without any departure

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-e7675af7 (Going-well departure with weak visible form: the hidden form is larges): the country 'GB' carries the effect (A/E 1.56) while the rest score IRE 0.96. O-11bad065
   • H-e7675af7 (Going-well departure with weak visible form: the hidden form is larges) passed discovery as a niche: A/E 1.35 on 94 bets (p 0.146), same-race control 0.95. O-83f2aaa3
   • H-9519a139 (Contrast: fell or unseated while NOT going well, at 14/1 or shorter, o) was predicted negative but came out the other way: A/E 1.13 on 367 bets (p 0.199 for the opposite direction). O-d93ddd07
2. AskedIs the weak-visible-form lift (best_last3_beaten > 5 at 14/1 or shorter) specific to horses that departed last time, or shared by any jumps runner with poor recent form at a short price? testing
3. MechanismA horse beaten more than five lengths in each of its last three runs, yet backed to 14/1 or shorter, is probably fancied for reasons outside its form figures: a drop in class, a change of trip or ground, or a yard's intent.
4. Why the market misses itThe form book shows only poor finishing positions, so casual money stays off. The informed money that shortens the price may not shorten it far enough.
5. Predictedpositive, A/E 1.05 to 1.12 on 1500+ bets if the rival holds
6. Falsified ifA/E 1.02 or below, which would leave the departure as the specific ingredient in H-e7675af7
7. Rule sketchbest_last3_beaten > 5 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15

Mechanism and rationale

A horse beaten more than five lengths in each of its last three runs, yet backed to 14/1 or shorter, is probably fancied for reasons outside its form figures: a drop in class, a change of trip or ground, or a yard's intent.

The rule

best_last3_beaten > 5 AND prior_runs >= 3 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15

Reads: best_last3_beaten, prior_runs, race_type, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

ae0bf5eb03c0d7e8214f58ed02e85bdaa2b5b6981009b6fc15b67bc0f69b2340

Frozen

not frozen

The definition can still change.

Confounders

• Class drops: a horse dropping in class is shortened for that reason, which the class-drop refinement isolates
• Trip or ground changes since the poor runs
• First-time headgear or a change of equipment
• Yard form and stable intent (strong yards' horses fancied despite poor form)
• Hidden departures: F/UR inside the last three runs give a large beaten figure or none at all; the no-departure refinement removes last-time departures
• Data window: best_last3_beaten starts 2025-11-01, so the whole sample comes from about one season and one set of going conditions
• Non-finishers' beaten distances may be missing, which could quietly change who meets the > 5 condition

Control

{
  "description": "The other runners in the same races should run at about A/E 0.97 to 1.00, the usual level. If they score as high as the selections, the lift belongs to the races and not to the poor-form angle. The main comparison for the question is the no-departure refinement against H-e7675af7 (departure plus weak form, 1.35 on 94). If the departure-free weak-form group at 14/1 or shorter reaches about 1.08 or better, the departure is not the special ingredient. If it is 1.02 or below, the departure is.",
  "rule": null
}

Designer

Discovery result

fail broad rule · 16 Oct, 13:40

Bets

8,301

1,101 winners

A/E

1.02

95%: 0.96 to 1.08

SP return

-17.1%

level stakes

Betfair SP return

+0.2%

8,295 bets with a BSP

Strike rate 13.3%. p = 0.23. 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

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 · 32s · 16 Oct, 13:36

  Output (schema-checked)

  {
    "rule": "best_last3_beaten > 5 AND prior_runs >= 3 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15",
    "refinements": [
      {
        "label": "no departure last time (clean rival)",
        "rule": "best_last3_beaten > 5 AND prior_runs >= 3 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15 AND COALESCE(last_fell_or_unseated, 0) = 0"
      },
      {
        "label": "class drop today (named reason outside the form figures)",
        "rule": "best_last3_beaten > 5 AND prior_runs >= 3 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15 AND race_class > last_race_class"
      }
    ],
    "control": {
      "description": "The other runners in the same races should run at about A/E 0.97 to 1.00, the usual level. If they score as high as the selections, the lift belongs to the races and not to the poor-form angle. The main comparison for the question is the no-departure refinement against H-e7675af7 (departure plus weak form, 1.35 on 94). If the departure-free weak-form group at 14/1 or shorter reaches about 1.08 or better, the departure is not the special ingredient. If it is 1.02 or below, the departure is.",
      "rule": null
    },
    "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": [
      "Class drops: a horse dropping in class is shortened for that reason, which the class-drop refinement isolates",
      "Trip or ground changes since the poor runs",
      "First-time headgear or a change of equipment",
      "Yard form and stable intent (strong yards' horses fancied despite poor form)",
      "Hidden departures: F/UR inside the last three runs give a large beaten figure or none at all; the no-departure refinement removes last-time departures",
      "Data window: best_last3_beaten starts 2025-11-01, so the whole sample comes from about one season and one set of going conditions",
      "Non-finishers' beaten distances may be missing, which could quietly change who meets the > 5 condition"
    ],
    "mechanism": "A jumps horse beaten more than five lengths in each of its last three runs (smallest beaten distance > 5 with at least three prior runs), yet backed to 14/1 or shorter, is fancied for reasons outside its form figures: a drop in class, a change of trip or ground, new equipment, or a yard's intent. If that is the source of H-e7675af7's 1.35, a fall or unseat last time is not needed.",
    "market_miss": "The form figures show only poor finishes, so casual and form-based money stays off. The informed money that brings the price into 14/1 or shorter may not bring it in far enough, which leaves the SP too long against the horse's true chance.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.05 to 1.12 on 1500+ bets if the rival holds",
      "falsified_if": "A/E 1.02 or below on the broad rule (or on the no-departure refinement). That would leave the departure as the specific ingredient in H-e7675af7. The rival also fails if the same-race control scores within 0.03 of the selections."
    },
    "notes": "I added prior_runs >= 3 so that 'beaten > 5 in each of the last three runs' holds literally, and is not met by horses with only one or two runs. The rule reads pre-off columns only and never the result. best_last3_beaten exists only from 2025-11; race_class for Ireland only from 2026-07, so the class-drop refinement is mostly GB. Both refinements follow from the question (rival without a departure) and from the mechanism (a class drop as the named outside reason)."
  }

Timeline

1. Job 1651 finished

   16 Oct, 13:40 · engine:compute-1

2. H-e857bd23 Discovery → KILLED fails: A/E 1.023, needs 1.05; p 0.2336, needs 0.05 or lower; periods [1.05, 0.98], each needs 1.0 (not a niche pass: A/E 1.023 below 1.05)

   16 Oct, 13:40 · engine:evaluator

3. discovery result for H-e857bd23 (broad rule): fail

   16 Oct, 13:40 · engine:evaluator

4. H-e857bd23 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 13:40 · engine:orchestrator

5. compute-1 took job 1651 (evaluate, attempt 1)

   16 Oct, 13:40 · compute-1

6. Job 1649 finished

   16 Oct, 13:37 · engine:ai-1

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

   16 Oct, 13:37 · engine:scheduler

8. H-e857bd23 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 13:37 · agent:designer@v1

9. ai-1 took job 1649 (design, attempt 1)

   16 Oct, 13:36 · ai-1

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

    16 Oct, 13:36 · engine:scheduler

11. New idea H-e857bd23: Rival: poor recent form at 14/1 or shorter beats its price without any departure (rival-explanation)

    16 Oct, 13:36 · 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.