Test-mechanism: the poor-jumping deficit should be absent in last-time winners whose jumping was good

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-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. AskedIn chases, do last-time winners with a top-third Jump Index score clearly higher A/E than those with a bottom-third index, so that poor jumping (not just having won) drives the deficit? answered
3. MechanismIf poor jumping is the cause, chase winners who jumped well last time carry no hidden flaw and should be priced fairly or better. That means A/E near or above the 1.07 control, and the gap to poor jumpers is the effect.
4. Why the market misses itThe market treats a last-time win the same way whatever the jumping quality, so good jumpers may even be slightly under-bet compared with poor jumpers at the same price.
5. Predictedpositive, A/E 1.00 to 1.12 on 300+ bets
6. Falsified ifA/E 0.85 or lower on 300+ bets, which would mean last-time chase winners are over-bet generally and jumping is not the cause
7. Rule sketchrace_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

Mechanism and rationale

If poor jumping is the cause, chase winners who jumped well last time carry no hidden flaw and should be priced fairly or better. That means A/E near or above the 1.07 control, and the gap to poor jumpers is the effect.

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 >= 0.67

Reads: race_type, last_position, last_race_type, last_jump_index_pct

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

bb688b6f260b04f1e3669a0a0961edbafe0b1ab3201709ac70f7c9324458d139

Frozen

not frozen

The definition can still change.

Confounders

• Good jumpers may simply be better horses and shorter in the betting. Compare the two groups within price bands, not just overall, because the poor-jumper deficit sat at 16/1+ and was at par at evens or shorter
• A rating rise after the win affects both groups. Check handicaps separately
• Course type (stiff versus sharp) may shape both the Jump Index last time and today's result
• Jump Index percentile depends on the field. A top-third score in a small or weak field may mean little
• Hurdle-to-chase switchers: a Jump Index earned over hurdles may not carry over to fences
• Coverage: Jump Index data is limited to tracked courses and dates, so the sample may lean toward particular tracks or periods

Control

{
  "description": "Same-race other runners: expected about 1.07, as in the poor-jumper study. The control that decides the mechanism is the mirror group: last-time chase winners with a bottom-third Jump Index (H-bfcb9669, A/E 0.78 on 240), measured over the same window and price bands. The mechanism holds if good jumpers score near or above 1.0 and clearly above the poor jumpers, especially at 9/1+. If both groups sit well below 1, the deficit belongs to last-time chase winners in general, not to poor 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 <= 0.33"
}

Discovery result

fail broad rule · 16 Oct, 16:53

Bets

902

212 winners

A/E

1.05

95%: 0.91 to 1.20

SP return

-10.4%

level stakes

Betfair SP return

+3.4%

880 bets with a BSP

Strike rate 23.5%. p = 0.26. Same-race control: the other runners in these races had A/E 0.98, 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 · 30s · 16 Oct, 16:51

  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 >= 0.67",
    "refinements": [
      {
        "label": "9/1 or longer (where the poor-jumper deficit sat)",
        "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 AND sp_decimal >= 10"
      },
      {
        "label": "Handicaps only (rating rise after the win hits both groups)",
        "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 AND is_handicap = 1"
      }
    ],
    "control": {
      "description": "Same-race other runners: expected about 1.07, as in the poor-jumper study. The control that decides the mechanism is the mirror group: last-time chase winners with a bottom-third Jump Index (H-bfcb9669, A/E 0.78 on 240), measured over the same window and price bands. The mechanism holds if good jumpers score near or above 1.0 and clearly above the poor jumpers, especially at 9/1+. If both groups sit well below 1, the deficit belongs to last-time chase winners in general, not to poor 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 <= 0.33"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Good jumpers may simply be better horses and shorter in the betting. Compare the two groups within price bands, not just overall, because the poor-jumper deficit sat at 16/1+ and was at par at evens or shorter",
      "A rating rise after the win affects both groups. Check handicaps separately",
      "Course type (stiff versus sharp) may shape both the Jump Index last time and today's result",
      "Jump Index percentile depends on the field. A top-third score in a small or weak field may mean little",
      "Hurdle-to-chase switchers: a Jump Index earned over hurdles may not carry over to fences",
      "Coverage: Jump Index data is limited to tracked courses and dates, so the sample may lean toward particular tracks or periods"
    ],
    "mechanism": "Poor jumping is a hidden flaw that a last-time win hides. A winner who jumped well carries no such flaw. If jumping is the cause of the poor-jumper deficit (A/E 0.78), chase runners who won last time with a top-third Jump Index should be priced fairly or slightly generously: A/E near or above the 1.07 same-race control. The gap between the good-jumper and poor-jumper groups is the size of the jumping effect.",
    "market_miss": "Punters and the odds-setters price a last-time win mostly on the result, not on how well the horse jumped. Jump Index sectional data is not widely used. So good and poor jumpers coming off a win are priced alike, and good jumpers may be slightly under-bet next to poor jumpers at the same price.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.00 to 1.12 on 300+ bets, clearly above the poor-jumper mirror group (0.78)",
      "falsified_if": "A/E 0.85 or lower on 300+ bets, or within 0.05 of the poor-jumper mirror group over the same window. Either would mean last-time chase winners are over-bet generally and poor jumping is not the cause."
    },
    "notes": "This is a mechanism test. Its main job is to answer Q-54be094c through the contrast with the poor-jumper group, not to find a bet on its own. A result of 1.00–1.05 would fail discovery yet still support the mechanism. Record it that way rather than as evidence against jumping. min_bets is raised to 300 to match the prediction's sample. The rule is unchanged from the sketch: every column is pre-off, and last_jump_index_pct is the same field used in H-bfcb9669, so the two groups are directly comparable. The 9/1+ refinement uses sp_decimal >= 10 to match the price zone where the poor-jumper deficit sat."
  }

Timeline

1. Job 2063 finished

   16 Oct, 16:53 · engine:compute-1

2. H-95f8568d Discovery → KILLED fails: A/E 1.046, needs 1.05; p 0.2634, needs 0.05 or lower; periods [1.08, 0.96], each needs 1.0 (not a niche pass: A/E 1.046 below 1.05)

   16 Oct, 16:53 · engine:evaluator

3. discovery result for H-95f8568d (broad rule): fail

   16 Oct, 16:53 · engine:evaluator

4. H-95f8568d Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:53 · engine:orchestrator

5. compute-1 took job 2063 (evaluate, attempt 1)

   16 Oct, 16:53 · compute-1

6. Job 2061 finished

   16 Oct, 16:51 · engine:ai-1

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

   16 Oct, 16:51 · engine:scheduler

8. H-95f8568d Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:51 · agent:designer@v1

9. ai-1 took job 2061 (design, attempt 1)

   16 Oct, 16:51 · ai-1

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

    16 Oct, 16:50 · engine:scheduler

11. New idea H-95f8568d: Test-mechanism: the poor-jumping deficit should be absent in last-time winners whose jumping was good (test-mechanism)

    16 Oct, 16:50 · 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.