Chase-bound winners that jumped poorly when winning 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-7dbfdd08 (Rival: big-rise quick-returning handicap winners stepping up in class ) was predicted negative but came out the other way: A/E 1.16 on 138 bets (p 0.225 for the opposite direction). O-23b5bd4e
   • H-21f9e6e7 (Long-priced raised handicap winners turned out within 14 days are over) pointed the predicted way but failed the bar: A/E 0.57 on 255 bets (p 0.967), control 1.03. O-7c9eecf8
2. AskedDo horses that won last time despite a poor Jump Index (bottom third of their field) run below their price today in a chase?
3. MechanismPoor jumping costs lengths and energy at every fence, and the risk of a fall grows as the fences and the opposition get stiffer. A horse that won while jumping badly won through the engine or a weak field. Next time, after a rise or against better rivals, those lost lengths decide the race or bring a mistake. It's a physical flaw that stays hidden behind a winning result.
4. Why the market misses itThe result shows a win, and race comments often call poor jumping 'novicey' and so forgivable. Bettors assume it will improve. The RaceiQ Jump Index shows how much ground was lost, but almost nobody prices it, so a winning horse's jumping is marked up by the result, not by the figure.
5. Predictednegative, A/E 0.80 to 0.92 on 300+ bets
6. Falsified ifA/E 0.97 or higher on 300+ bets, or no lower than horses that won last time with a Jump Index in the top third.
7. Rule sketchrace_type = 'Chase' AND last_position = 1 AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33

Mechanism and rationale

Poor jumping costs lengths and energy at every fence, and the risk of a fall grows as the fences and the opposition get stiffer. A horse that won while jumping badly won through the engine or a weak field. Next time, after a rise or against better rivals, those lost lengths decide the race or bring a mistake. It's a physical flaw that stays hidden behind a winning result.

The rule

race_type = 'Chase' AND last_position = 1 AND last_race_type IN ('Chase', 'Hurdle') AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33

Reads: race_type, last_position, last_race_type, last_jump_index_pct

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

3a6b9d0ad6d1f7619b50ca9baef8f548a48381d75045211da1ef8c0afcf70642

Frozen

not frozen

The definition can still change.

Confounders

• The last run may have been a hurdle. Hurdle jumping is a weaker guide to fence jumping, so it is tested separately (refinement 1).
• Jump Index only exists from 2023-03 in GB and 2024-01 in IRE. The sample is mostly GB and recent, so read it by country and by period.
• In small last-time fields (2 to 5 runners) the percentile is coarse, and a field with one other runner forces 0 or 1 (refinement 2).
• Novice versus handicap chases: poor jumping in novices is often forgiven and improves. In handicaps the horse also carries a rise from the handicapper (refinement 3).
• Last-time winners as a group may be over-bet whatever their jumping. The top-third Jump Index winners control separates that out.
• Overlaps NEG-repeat-jumping-errors, where the race comments also flag errors. Read the two forward records together, not as independent evidence.
• Class rise or a step up in opposition after the win can produce the effect without the jumping. Check the result within class-rise and same-class groups.

Control

{
  "description": "Same-race other runners should sit near A/E 1.0, which shows the race itself is not mispriced. The control that matters is last-time winners who jumped well: chase runners today that won last time with a Jump Index in the top third of their field. The prediction is that they price fairly (A/E about 0.95 to 1.05). The poor-jumping winners must come out clearly below them, about 0.08 or more lower in A/E. If both groups are equally low, the effect comes from last-time winners in general being over-bet, not from the jumping.",
  "rule": "race_type = 'Chase' AND last_position = 1 AND last_race_type IN ('Chase', 'Hurdle') AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct >= 0.67"
}

Discovery result

fail broad rule · 16 Oct, 16:47

Bets

240

39 winners

A/E

0.78

95%: 0.56 to 1.07

SP return

-49.8%

level stakes

Betfair SP return

-45.8%

234 bets with a BSP

Strike rate 16.2%. p = 0.95. Same-race control: the other runners in these races had A/E 1.07, 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 · 44s · 16 Oct, 16:47

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND last_position = 1 AND last_race_type IN ('Chase', 'Hurdle') AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33",
    "refinements": [
      {
        "label": "won a chase last time (fence jumping only)",
        "rule": "race_type = 'Chase' AND last_position = 1 AND last_race_type = 'Chase' AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33"
      },
      {
        "label": "last field of 6+ so the percentile is not coarse",
        "rule": "race_type = 'Chase' AND last_position = 1 AND last_race_type IN ('Chase', 'Hurdle') AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33 AND last_field_size >= 6"
      },
      {
        "label": "handicap chase today (carrying a rise for the win)",
        "rule": "race_type = 'Chase' AND last_position = 1 AND last_race_type IN ('Chase', 'Hurdle') AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct <= 0.33 AND is_handicap = 1"
      }
    ],
    "control": {
      "description": "Same-race other runners should sit near A/E 1.0, which shows the race itself is not mispriced. The control that matters is last-time winners who jumped well: chase runners today that won last time with a Jump Index in the top third of their field. The prediction is that they price fairly (A/E about 0.95 to 1.05). The poor-jumping winners must come out clearly below them, about 0.08 or more lower in A/E. If both groups are equally low, the effect comes from last-time winners in general being over-bet, not from the jumping.",
      "rule": "race_type = 'Chase' AND last_position = 1 AND last_race_type IN ('Chase', 'Hurdle') AND last_jump_index_pct IS NOT NULL AND last_jump_index_pct >= 0.67"
    },
    "pass_criteria": {
      "min_bets": 200,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "The last run may have been a hurdle. Hurdle jumping is a weaker guide to fence jumping, so it is tested separately (refinement 1).",
      "Jump Index only exists from 2023-03 in GB and 2024-01 in IRE. The sample is mostly GB and recent, so read it by country and by period.",
      "In small last-time fields (2 to 5 runners) the percentile is coarse, and a field with one other runner forces 0 or 1 (refinement 2).",
      "Novice versus handicap chases: poor jumping in novices is often forgiven and improves. In handicaps the horse also carries a rise from the handicapper (refinement 3).",
      "Last-time winners as a group may be over-bet whatever their jumping. The top-third Jump Index winners control separates that out.",
      "Overlaps NEG-repeat-jumping-errors, where the race comments also flag errors. Read the two forward records together, not as independent evidence.",
      "Class rise or a step up in opposition after the win can produce the effect without the jumping. Check the result within class-rise and same-class groups."
    ],
    "mechanism": "Poor jumping costs lengths and energy at every fence, and the risk of a mistake or fall grows as the fences and the opposition get stiffer. A horse that won while in the bottom third of its field on the RaceiQ Jump Index won through its engine or a weak field. Next time, after a handicap rise or against better rivals, those lost lengths decide the race or bring a mistake. The flaw is physical and the winning result hides it.",
    "market_miss": "The form figure shows a 1. Race comments call poor jumping 'novicey' and forgivable, and bettors assume it will improve. The Jump Index measures the ground the horse lost at its fences, but it is barely used in pricing. So the market rates the horse on the result rather than on how much ground its jumping cost, and its price today is too short.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.80 to 0.92 on 300+ bets, at least 0.08 below the top-third Jump Index last-time-winner control",
      "falsified_if": "A/E 0.97 or higher on 300+ bets, or an A/E no lower than the control (chase runners that won last time with a top-third Jump Index)."
    },
    "notes": "I checked the catalogue: last_jump_index_pct is the share of the rest of its last field the horse beat on Jump Index (0 to 1), so <= 0.33 is the bottom third. last_race_type IN ('Chase','Hurdle') makes the jumps-only intent explicit and stops a Flat or Bumper run from leaking in. min_bets is raised to 200 because the prediction needs 300+ bets and Jump Index coverage is short."
  }

Timeline

1. Job 2054 finished

   16 Oct, 16:47 · engine:compute-1

2. H-bfcb9669 Discovery → KILLED fails: 240 bets, needs 300; p 0.0668 (lower tail), needs 0.05

   16 Oct, 16:47 · engine:evaluator

3. discovery result for H-bfcb9669 (broad rule): fail

   16 Oct, 16:47 · engine:evaluator

4. H-bfcb9669 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:47 · engine:orchestrator

5. compute-1 took job 2054 (evaluate, attempt 1)

   16 Oct, 16:47 · compute-1

6. Job 2051 finished

   16 Oct, 16:47 · engine:ai-1

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

   16 Oct, 16:47 · engine:scheduler

8. H-bfcb9669 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:47 · agent:designer@v1

9. ai-1 took job 2051 (design, attempt 1)

   16 Oct, 16:47 · ai-1

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

    16 Oct, 16:45 · engine:scheduler

11. New idea H-bfcb9669: Chase-bound winners that jumped poorly when winning are over-bet (scout)

    16 Oct, 16:45 · agent:scout@v1

Every change is an event with an actor: an agent and its version, an engine, a human, or the importer. Current stage: Killed.