Replicate: momentum interrupted, beaten 2 lengths or less, in hurdles and chases

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
   • H19-momentum-interrupted-beaten-narrowly (H19: momentum interrupted beaten narrowly) was GREEN in the earlier research: A/E 1.18 on 242 bets (periods 1.07 / 1.26); now PROMISING. O-a6c61d14
2. AskedDoes the momentum-interrupted, narrowly-beaten lift hold in hurdles and chases, or only on the Flat where discovery was strongest? testing
3. MechanismA horse checked while making its move loses momentum it cannot fully win back. Over jumps a check is less decisive than on the Flat, but a horse that still finished within 2 lengths should still have more ability than the result shows.
4. Why the market misses itThe result shows a beaten horse. Jumps punters weigh jumping errors more than checks in running, so a check may be under-read.
5. Predictedpositive, A/E 1.05 to 1.15 on 150+ bets
6. Falsified ifA/E below 1.0 on 150+ bets, or a lift in only one period
7. Rule sketchlast_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase')
8. question id

Mechanism and rationale

A horse checked while making its move loses momentum it cannot fully win back. Over jumps a check is less decisive than on the Flat, but a horse that still finished within 2 lengths should still have more ability than the result shows.

The rule

last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase')

Reads: last_momentum_interrupted, last_position, last_beaten_lengths, race_type

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

46130758975fbbc83d3104eff992c014f254e45c6c6b6db849e1047d276eea57

Frozen

not frozen

The definition can still change.

Confounders

• Small sample: last_momentum_interrupted exists only from 2025-11-01, so discovery covers about one jumps season
• Jumping errors recorded as or alongside interruption (a blunder that stops a run); refinement 2 separates them
• Narrow beaten margin alone may carry the lift (control: uninterrupted close losers)
• Going: interruptions and close finishes are more common on soft ground in big fields
• Code switchers: today's jumps runner may have been checked on the Flat or in a bumper (refinement 1)
• Overlap with H19, which already includes jumps runners, so this is not independent of the parent result

Control

{
  "description": "Other runners in the same jumps races should sit near par (A/E ~0.95-1.0), showing the lift is not a property of the races. Secondary comparison: jumps runners beaten 2 lengths or less last time with no interruption, to separate the check from simply being narrowly beaten (the rival hypothesis H-66cafe6f). If that group scores as well, the 'check' adds nothing.",
  "rule": "last_momentum_interrupted = 0 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase')"
}

Designer

designer@v1

Direction

Discovery result

fail broad rule · 16 Oct, 14:46

Bets

22

1 winners

A/E

0.27

95%: 0.01 to 1.51

SP return

-31.8%

level stakes

Betfair SP return

–

22 bets with a BSP

Strike rate 4.5%. p = 0.98. Same-race control: the other runners in these races had A/E 1.15, 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 · 31s · 16 Oct, 14:45

  Output (schema-checked)

  {
    "rule": "last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase')",
    "refinements": [
      {
        "label": "check came over jumps too",
        "rule": "last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase') AND last_race_type IN ('Hurdle','Chase')"
      },
      {
        "label": "interruption not a jumping error",
        "rule": "last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase') AND last_jumping_error = 0"
      },
      {
        "label": "14/1 or shorter",
        "rule": "last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase') AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "Other runners in the same jumps races should sit near par (A/E ~0.95-1.0), showing the lift is not a property of the races. Secondary comparison: jumps runners beaten 2 lengths or less last time with no interruption, to separate the check from simply being narrowly beaten (the rival hypothesis H-66cafe6f). If that group scores as well, the 'check' adds nothing.",
      "rule": "last_momentum_interrupted = 0 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase')"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Small sample: last_momentum_interrupted exists only from 2025-11-01, so discovery covers about one jumps season",
      "Jumping errors recorded as or alongside interruption (a blunder that stops a run); refinement 2 separates them",
      "Narrow beaten margin alone may carry the lift (control: uninterrupted close losers)",
      "Going: interruptions and close finishes are more common on soft ground in big fields",
      "Code switchers: today's jumps runner may have been checked on the Flat or in a bumper (refinement 1)",
      "Overlap with H19, which already includes jumps runners, so this is not independent of the parent result"
    ],
    "mechanism": "A horse checked while making its move loses momentum it cannot fully win back. A jumps horse that was checked yet still finished within 2 lengths ran better than its placing and margin show, so its true ability is understated by the form figure.",
    "market_miss": "The form figure shows a beaten horse and the margin looks ordinary. Jumps punters and compilers weigh jumping errors heavily but read in-running checks lightly, so the hidden merit of a checked close loser is under-priced.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.05 to 1.15 on 150+ bets",
      "falsified_if": "A/E below 1.0 on 150+ bets; or the lift appears in only one period; or uninterrupted close losers over jumps score as well or better (no added value from the check)."
    },
    "notes": "Replication of H19 limited to jumps races today. Broad rule kept as sketched; race_type is today's race. The momentum column only exists from 2025-11-01, so the sample will be thin and probably mostly one season; a thin result pointing the right way goes on to falsification and forward tracking rather than being killed. The findings say the parent edge lives at BSP more than SP, so judge forward at exchange prices."
  }

Timeline

1. Job 1834 finished

   16 Oct, 14:46 · engine:compute-1

2. H-b827bdc5 Discovery → KILLED fails: 22 bets, needs 150; A/E 0.271, needs 1.05; p 0.9752, needs 0.05 or lower; periods [], each needs 1.0; A/E 0.271 against same-race control 1.152, needs 0.03 above (not a niche pass: only 22 bets; A/E 0.271 below 1.05; p 0.9751773813715859 above 0.5; does not beat the same-race control 1.152)

   16 Oct, 14:46 · engine:evaluator

3. discovery result for H-b827bdc5 (broad rule): fail

   16 Oct, 14:46 · engine:evaluator

4. H-b827bdc5 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 14:46 · engine:orchestrator

5. compute-1 took job 1834 (evaluate, attempt 1)

   16 Oct, 14:46 · compute-1

6. Job 1831 finished

   16 Oct, 14:46 · engine:ai-1

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

   16 Oct, 14:46 · engine:scheduler

8. H-b827bdc5 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 14:46 · agent:designer@v1

9. ai-1 took job 1831 (design, attempt 1)

   16 Oct, 14:45 · ai-1

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

    16 Oct, 14:45 · engine:scheduler

11. New idea H-b827bdc5: Replicate: momentum interrupted, beaten 2 lengths or less, in hurdles and chases (replicate)

    16 Oct, 14:45 · 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.