Rival: closers that made late headway and were beaten narrowly without trouble

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-48a2eca6 (Test-mechanism: the edge sits with horses the market did not shorten a) passed discovery as a niche: A/E 1.24 on 104 bets (p 0.25), same-race control 0.97. O-22d00ba5
2. AskedIs the lift really hold-up/closing run style, rather than trouble? Horses that made headway late and were beaten narrowly without trouble would test this. answered
3. MechanismHorses that get checked are mostly held up and closing. If closers beaten narrowly are under-priced in general, H19 would be measuring run style rather than the interruption.
4. Why the market misses itIf the rival is true, punters over-weight the beaten margin and under-weight a closing run against the pace.
5. Predictedpositive, If the rival is true, A/E 1.10+ on 1000+ bets; expected around 1.0
6. Falsified ifA/E at or below 1.03 on 1000+ bets. That rules out the rival and leaves the trouble flag as the source.
7. Rule sketchlast_momentum_interrupted = 0 AND last_made_headway = 1 AND last_position > 1 AND last_beaten_lengths <= 2

Mechanism and rationale

Horses that get checked are mostly held up and closing. If closers beaten narrowly are under-priced in general, H19 would be measuring run style rather than the interruption.

The rule

last_momentum_interrupted = 0 AND last_made_headway = 1 AND last_position > 1 AND last_beaten_lengths <= 2

Reads: last_momentum_interrupted, last_made_headway, last_position, last_beaten_lengths

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

733c699614a682e2253e69521aebc32ea3b7e1cbb559fd2159cb97d8cee3af36

Frozen

not frozen

The definition can still change.

Confounders

• pace of the last race: closers flatter themselves behind a strong pace that collapsed
• course shape and track pace bias favouring closers last time
• how consistently the comment writer flags 'headway' compared with 'checked' or 'interrupted', and whether the two flags overlap
• run-style mix: held-up horses are more common in H19 than in the general narrowly beaten pool
• price band: any lift could sit in one short-price band, as in earlier families

Control

{
  "description": "The same-race other runners should be at about par (A/E ~1.0). The key comparison is narrowly beaten runners with no trouble who did NOT make late headway. With the H-66cafe6f baseline at 0.98 on 12,205, if closing is what carries the lift, the headway group should clearly beat this control. If both sit near 1.0, run style does not explain H19's 1.18, and the interruption flag stays the source.",
  "rule": "last_momentum_interrupted = 0 AND last_made_headway = 0 AND last_position > 1 AND last_beaten_lengths <= 2"
}

Designer

designer@v1

Direction

positive

Evaluator version

eval-1.0

Discovery result

fail broad rule · 16 Oct, 14:49

Bets

5,540

954 winners

A/E

0.97

95%: 0.91 to 1.03

SP return

-20.8%

level stakes

Betfair SP return

-4.6%

5,515 bets with a BSP

Strike rate 17.2%. p = 0.84. Same-race control: the other runners in these races had A/E 1.01, so the selection did not 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 · 25s · 16 Oct, 14:48

  Output (schema-checked)

  {
    "rule": "last_momentum_interrupted = 0 AND last_made_headway = 1 AND last_position > 1 AND last_beaten_lengths <= 2",
    "refinements": [
      {
        "label": "held-up or midfield closers (matches H19's run-style mix)",
        "rule": "last_momentum_interrupted = 0 AND last_made_headway = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND last_run_style IN ('held_up', 'midfield')"
      },
      {
        "label": "14/1 or shorter today",
        "rule": "last_momentum_interrupted = 0 AND last_made_headway = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "The same-race other runners should be at about par (A/E ~1.0). The key comparison is narrowly beaten runners with no trouble who did NOT make late headway. With the H-66cafe6f baseline at 0.98 on 12,205, if closing is what carries the lift, the headway group should clearly beat this control. If both sit near 1.0, run style does not explain H19's 1.18, and the interruption flag stays the source.",
      "rule": "last_momentum_interrupted = 0 AND last_made_headway = 0 AND last_position > 1 AND last_beaten_lengths <= 2"
    },
    "pass_criteria": {
      "min_bets": 1000,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 200,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "pace of the last race: closers flatter themselves behind a strong pace that collapsed",
      "course shape and track pace bias favouring closers last time",
      "how consistently the comment writer flags 'headway' compared with 'checked' or 'interrupted', and whether the two flags overlap",
      "run-style mix: held-up horses are more common in H19 than in the general narrowly beaten pool",
      "price band: any lift could sit in one short-price band, as in earlier families"
    ],
    "mechanism": "Horses that get checked are mostly held up and closing. If narrowly beaten closers are under-priced in general, H19's lift (1.18 on 242) would reflect run style rather than the interruption. This rule isolates closing headway with no trouble, so it tests whether closing alone carries the edge.",
    "market_miss": "If the rival is true, punters over-weight the bare beaten margin and under-weight a closing run made against the run of the race, so a narrow defeat after late headway is priced like any other narrow defeat.",
    "prediction": {
      "direction": "positive",
      "size": "If the rival is true, A/E 1.10+ on 1000+ bets. The prior expectation is about 1.0, close to the 0.98 no-trouble baseline.",
      "falsified_if": "A/E at or below 1.03 on 1000+ bets, or not clearly above the no-headway control. That rules out the rival and leaves the momentum-interrupted flag as the source of H19's lift."
    },
    "notes": "This is a rival-explanation test for H19 / Q-e3059732. A null result is informative and expected; it strengthens H19. The rule sketch was kept unchanged: every column is a pre-off last-run field, and last_position > 1 excludes non-finishers (NULL). The refinements are read only if the broad rule shows evidence."
  }

Timeline

1. Job 1843 finished

   16 Oct, 14:49 · engine:compute-1

2. H-fe5384b6 Discovery → KILLED fails: A/E 0.969, needs 1.05; p 0.8358, needs 0.05 or lower; periods [0.98, 0.97], each needs 1.0; A/E 0.969 against same-race control 1.009, needs 0.03 above (not a niche pass: A/E 0.969 below 1.05; p 0.8357858140525269 above 0.5; does not beat the same-race control 1.009)

   16 Oct, 14:49 · engine:evaluator

3. discovery result for H-fe5384b6 (broad rule): fail

   16 Oct, 14:49 · engine:evaluator

4. H-fe5384b6 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 14:49 · engine:orchestrator

5. compute-1 took job 1843 (evaluate, attempt 1)

   16 Oct, 14:49 · compute-1

6. Job 1840 finished

   16 Oct, 14:48 · engine:ai-1

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

   16 Oct, 14:48 · engine:scheduler

8. H-fe5384b6 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 14:48 · agent:designer@v1

9. ai-1 took job 1840 (design, attempt 1)

   16 Oct, 14:48 · ai-1

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

    16 Oct, 14:47 · engine:scheduler

11. New idea H-fe5384b6: Rival: closers that made late headway and were beaten narrowly without trouble (rival-explanation)

    16 Oct, 14:47 · 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.