Sharpen: poor hurdle jumpers placed last time, now running over fences, are over-bet

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-95f8568d (Test-mechanism: the poor-jumping deficit should be absent in last-time) pointed the predicted way but failed the bar: A/E 1.05 on 902 bets (p 0.263), control 0.98. O-205a18e1
   • H-bfcb9669 (Chase-bound winners that jumped poorly when winning are over-bet) pointed the predicted way but failed the bar: A/E 0.78 on 240 bets (p 0.951), control 1.07. O-7acbc407
2. AskedIs the poor-jumper deficit stronger for horses whose last run was over hurdles (first chase or early chase career) than for those already chasing? answered
3. MechanismFences punish poor jumping far more than hurdles do. A horse that jumped hurdles poorly but finished close up is the most exposed when it steps up to fences, and has no chase form to give it away.
4. Why the market misses itThe market prices a recent hurdle placing as a sign of ability. It has no chase jumping evidence yet, and the poor hurdle Jump Index is not in the form book.
5. Predictednegative, A/E 0.80 to 0.92 on 200+ bets
6. Falsified ifA/E 0.97 or higher on 200+ bets, or no lower than the same pool with last_jump_index_pct >= 0.67
7. Rule sketchrace_type = 'Chase' AND last_race_type = 'Hurdle' AND last_position <= 3 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33
8. question id

The question's answer so far: No. Poor hurdle jumpers stepping up to fences scored 0.83 on 150, no stronger than the parent's 0.78. The deficit sits entirely in Ireland (0.44 on 57), while GB was at par (1.03 on 93).

Mechanism and rationale

Fences punish poor jumping far more than hurdles do. A horse that jumped hurdles poorly but finished close up is the most exposed when it steps up to fences, and has no chase form to give it away.

The rule

race_type = 'Chase' AND last_race_type = 'Hurdle' AND last_position <= 3 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33

Reads: race_type, last_race_type, last_position, last_jump_index_pct

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

aaabcab8e6b0341659dc2e46aa28c94b2830ac6b173aa034304fa5e99c843974

Frozen

not frozen

The definition can still change.

Confounders

• Novice chase class: most hurdle-to-fence switchers run in novice or beginners' chases, where everyone's jumping is untested. The good-jumper mirror pool controls for this.
• Stable schooling quality: yards that school well over fences may also have horses with better hurdle Jump Index, mixing yard effects with horse effects
• Field size and fall or unseat rates in novice chases, which hit long-priced runners hardest
• Price-band effect: earlier poor-jumper deficits sat at 16/1+, so a broad result may be favourite-longshot bias. Check it against the same-race control by price band
• Country imbalance: the earlier Irish poor-jumper result (0.45) rested on a tiny sample. Check GB and IRE separately
• Sample size: placed hurdlers who are poor jumpers and go chasing may give fewer than 200 bets

Control

{
  "description": "The same-race control (other runners in the same chases) should be close to par, about 0.98 to 1.07, as in the earlier chase jumping tests. The key comparison is the mirror pool: hurdle-placed horses now over fences whose last Jump Index was in the top third. If poor jumping is the cause, that pool should score clearly higher (near 1.0 or above), and the gap should be about 0.1 A/E points or more. If the mirror pool is just as low, the deficit belongs to hurdle-to-chase switchers in general (novice chase class, field size), not to jumping.",
  "rule": "race_type = 'Chase' AND last_race_type = 'Hurdle' AND last_position <= 3 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct >= 0.67"
}

Discovery result

fail broad rule · 16 Oct, 16:56

Bets

150

22 winners

A/E

0.83

95%: 0.52 to 1.25

SP return

-48.3%

level stakes

Betfair SP return

-42.5%

142 bets with a BSP

Strike rate 14.7%. p = 0.84. Same-race control: the other runners in these races had A/E 1.05, 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 · 34s · 16 Oct, 16:55

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND last_race_type = 'Hurdle' AND last_position <= 3 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33",
    "refinements": [
      {
        "label": "First-time chasers only (no chase form to expose the jumping)",
        "rule": "race_type = 'Chase' AND last_race_type = 'Hurdle' AND last_position <= 3 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33 AND chase_runs IS NOT NULL AND chase_runs = 0"
      },
      {
        "label": "Longer prices (the poor-jumper deficit sat at 16/1+, short-priced class horses survived)",
        "rule": "race_type = 'Chase' AND last_race_type = 'Hurdle' AND last_position <= 3 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33 AND sp_decimal > 9"
      }
    ],
    "control": {
      "description": "The same-race control (other runners in the same chases) should be close to par, about 0.98 to 1.07, as in the earlier chase jumping tests. The key comparison is the mirror pool: hurdle-placed horses now over fences whose last Jump Index was in the top third. If poor jumping is the cause, that pool should score clearly higher (near 1.0 or above), and the gap should be about 0.1 A/E points or more. If the mirror pool is just as low, the deficit belongs to hurdle-to-chase switchers in general (novice chase class, field size), not to jumping.",
      "rule": "race_type = 'Chase' AND last_race_type = 'Hurdle' AND last_position <= 3 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct >= 0.67"
    },
    "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": [
      "Novice chase class: most hurdle-to-fence switchers run in novice or beginners' chases, where everyone's jumping is untested. The good-jumper mirror pool controls for this.",
      "Stable schooling quality: yards that school well over fences may also have horses with better hurdle Jump Index, mixing yard effects with horse effects",
      "Field size and fall or unseat rates in novice chases, which hit long-priced runners hardest",
      "Price-band effect: earlier poor-jumper deficits sat at 16/1+, so a broad result may be favourite-longshot bias. Check it against the same-race control by price band",
      "Country imbalance: the earlier Irish poor-jumper result (0.45) rested on a tiny sample. Check GB and IRE separately",
      "Sample size: placed hurdlers who are poor jumpers and go chasing may give fewer than 200 bets"
    ],
    "mechanism": "Fences punish poor jumping far more than hurdles do: a mistake at a fence costs more ground and more often ends in a fall or unseat. A horse that finished in the first three over hurdles while jumping poorly (bottom-third Jump Index) got away with it over the smaller obstacle. When it steps up to fences that flaw is exposed, so it should run below the ability its hurdle placing suggests.",
    "market_miss": "The market prices a recent hurdle placing as proof of ability, and for a horse switching codes there is no chase jumping evidence to mark it down. The Jump Index is a sectional or tracking measure that is not in the form book, so a placed but poor-jumping hurdler looks the same as a slick one to most punters.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.80 to 0.92 on 200+ bets, against a same-race control near 1.0",
      "falsified_if": "A/E 0.97 or higher on 200+ bets, or the broad pool's A/E no lower than the mirror pool (same rule with last_jump_index_pct >= 0.67)"
    },
    "notes": "I kept the rule sketch as it is. All of its columns are pre-off, it does not read the result, and it slices only what the mechanism needs (fence race, last run over hurdles, placed, poor Jump Index). The refinements come from the reasoning: a first-time chaser has no chase form, and the earlier deficit sat at longer prices. The direction is negative, so the vault judges the rule as over-bet. The floors are taken unchanged, measured in the predicted direction."
  }

Timeline

1. Job 2073 finished

   16 Oct, 16:56 · engine:compute-1

2. H-730c9035 Discovery → KILLED fails: 150 bets, needs 300; p 0.2193 (lower tail), needs 0.05

   16 Oct, 16:56 · engine:evaluator

3. discovery result for H-730c9035 (broad rule): fail

   16 Oct, 16:56 · engine:evaluator

4. H-730c9035 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:56 · engine:orchestrator

5. compute-1 took job 2073 (evaluate, attempt 1)

   16 Oct, 16:56 · compute-1

6. Job 2071 finished

   16 Oct, 16:56 · engine:ai-1

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

   16 Oct, 16:56 · engine:scheduler

8. H-730c9035 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:56 · agent:designer@v1

9. ai-1 took job 2071 (design, attempt 1)

   16 Oct, 16:55 · ai-1

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

    16 Oct, 16:55 · engine:scheduler

11. New idea H-730c9035: Sharpen: poor hurdle jumpers placed last time, now running over fences, are over-bet (sharpen)

    16 Oct, 16:55 · 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.