Exposed young marathon handicap chasers sit at par: the lift needs an unexposed chaser

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-aa66431f (Young marathon handicap chasers with little chase experience carry the) passed discovery as a niche: A/E 1.24 on 75 bets (p 0.282), same-race control 0.93. O-eba48f81
2. AskedIs the young marathon-chase lift specific to horses still unexposed over fences, whose chase form understates how much they will improve at an extreme trip? testing
3. MechanismA young horse with few chase runs has jumping and stamina still developing, so its handicap mark and form lag its ability at an extreme trip. A young horse with eight or more chase runs has been fully assessed by the handicapper, so that hidden improvement is gone.
4. Why the market misses itBettors and the handicapper price unexposed chasers off a short, low-trip chase record that understates them. For exposed horses the form book is complete, so their price should be fair.
5. Predictednegative, A/E 0.90 to 1.02 on 150+ bets, clearly below the 1.24 of the unexposed slice
6. Falsified ifA/E at or above 1.12 on 150+ bets, which would mean experience does not matter and the unexposed-chaser mechanism is wrong
7. Rule sketchrace_type = 'Chase' AND is_handicap = 1 AND age <= 7 AND distance_yards >= 6160 AND chase_runs >= 8
8. question idQ-0c77086a

Mechanism and rationale

A young horse with few chase runs has jumping and stamina still developing, so its handicap mark and form lag its ability at an extreme trip. A young horse with eight or more chase runs has been fully assessed by the handicapper, so that hidden improvement is gone.

The rule

race_type = 'Chase' AND is_handicap = 1 AND age <= 7 AND distance_yards >= 6160 AND prior_chase_runs >= 8

Reads: race_type, is_handicap, age, distance_yards, prior_chase_runs

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

30ea6469a4618ee45e0bb6d2c998de57ebef4047e3259de20f91d5dc81837209

Frozen

not frozen

The definition can still change.

Confounders

• Class: exposed young chasers with 8+ chase runs by age 7 tend to be higher-rated and run in better marathon handicaps (Nationals, festival handicaps), so any gap against the unexposed group may reflect class or race tier rather than exposure
• Short-price cluster: every earlier positive marathon slice was carried by the up-to-7/1 band alone (1.33 to 1.68, rest 0.89 to 0.95); whichever exposure group holds more short-priced runners will look better for that reason alone (checked by refinement 1)
• Small samples at 3m4f+: the unexposed contrast group is only about 75 bets, so the gap between groups will have a wide interval
• Age within the band: horses with 8+ chase runs are mostly 7-year-olds, while the unexposed group skews 5 to 6, which mixes age with exposure
• Coverage: chase-run counts start in August 2020, so early-window horses can look less exposed than they really are, which may push truly exposed horses into the unexposed group

Control

{
  "description": "Two comparisons. (1) Standard same-race control: the other runners in these races; expected near par (0.90 to 1.00, as in earlier slices at 0.85 to 0.93), so an exposed-group reading close to that control supports 'no lift'. (2) The mechanism's contrast group: young marathon handicap chasers with 6 or fewer prior chase runs, measured over the same window, which earlier scored 1.24 on 75. The hypothesis is supported only if the exposed group sits clearly below this unexposed group; if both read alike (both high or both par), experience is not the lever. Runners with exactly 7 prior chase runs are left out of both groups as a buffer.",
  "rule": "race_type = 'Chase' AND is_handicap = 1 AND age <= 7 AND distance_yards >= 6160 AND prior_chase_runs <= 6"
}

Discovery result

fail broad rule · 16 Oct, 14:39

Bets

33

3 winners

A/E

0.84

95%: 0.17 to 2.45

SP return

-55.1%

level stakes

Betfair SP return

-52.0%

33 bets with a BSP

Strike rate 9.1%. p = 0.69. Same-race control: the other runners in these races had A/E 1.04, 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 · 42s · 16 Oct, 14:38

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND is_handicap = 1 AND age <= 7 AND distance_yards >= 6160 AND prior_chase_runs >= 8",
    "refinements": [
      {
        "label": "Short-price band among exposed young marathon chasers (tests whether the up-to-7/1 cluster that carried every earlier positive slice sits in the exposed group)",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND age <= 7 AND distance_yards >= 6160 AND prior_chase_runs >= 8 AND sp_decimal <= 8.0"
      },
      {
        "label": "Dose-response: heavily exposed young chasers (12+ chase runs) should sit lowest if exposure removes the hidden improvement",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND age <= 7 AND distance_yards >= 6160 AND prior_chase_runs >= 12"
      }
    ],
    "control": {
      "description": "Two comparisons. (1) Standard same-race control: the other runners in these races; expected near par (0.90 to 1.00, as in earlier slices at 0.85 to 0.93), so an exposed-group reading close to that control supports 'no lift'. (2) The mechanism's contrast group: young marathon handicap chasers with 6 or fewer prior chase runs, measured over the same window, which earlier scored 1.24 on 75. The hypothesis is supported only if the exposed group sits clearly below this unexposed group; if both read alike (both high or both par), experience is not the lever. Runners with exactly 7 prior chase runs are left out of both groups as a buffer.",
      "rule": "race_type = 'Chase' AND is_handicap = 1 AND age <= 7 AND distance_yards >= 6160 AND prior_chase_runs <= 6"
    },
    "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": [
      "Class: exposed young chasers with 8+ chase runs by age 7 tend to be higher-rated and run in better marathon handicaps (Nationals, festival handicaps), so any gap against the unexposed group may reflect class or race tier rather than exposure",
      "Short-price cluster: every earlier positive marathon slice was carried by the up-to-7/1 band alone (1.33 to 1.68, rest 0.89 to 0.95); whichever exposure group holds more short-priced runners will look better for that reason alone (checked by refinement 1)",
      "Small samples at 3m4f+: the unexposed contrast group is only about 75 bets, so the gap between groups will have a wide interval",
      "Age within the band: horses with 8+ chase runs are mostly 7-year-olds, while the unexposed group skews 5 to 6, which mixes age with exposure",
      "Coverage: chase-run counts start in August 2020, so early-window horses can look less exposed than they really are, which may push truly exposed horses into the unexposed group"
    ],
    "mechanism": "A young (7 or under) chaser with few chase runs is still improving its jumping and stamina, so at 3m4f+ its handicap mark and visible form trail its true ability. A young chaser with 8+ prior chase runs has been fully assessed by the handicapper and the form book, so that hidden improvement is gone and it should run to its price. If the young marathon lift (1.16 on 618 overall, 1.24 on 75 for the unexposed) comes from unexposed improvement, the exposed group should lose it.",
    "market_miss": "For unexposed chasers, bettors and the handicapper work from a short chase record, mostly at shorter trips, that understates them at an extreme trip. For exposed young chasers the record is complete and includes long-trip runs, so there is nothing hidden to misprice and the SP should be fair or slightly generous to them because of their reputation.",
    "prediction": {
      "direction": "negative",
      "size": "A/E about 0.90 to 1.02 on 150+ bets: at par or slightly over-bet, and clearly below the unexposed young group (1.24 on 75)",
      "falsified_if": "A/E at or above 1.12 on 150+ bets, or the exposed group matching or beating the unexposed control group (prior_chase_runs <= 6). Either would mean chase experience does not remove the lift and the unexposed-chaser mechanism is wrong."
    },
    "notes": "I swapped chase_runs for prior_chase_runs. The catalogue describes prior_chase_runs as the record before today, and it is the same column as the unexposed slice H-aa66431f (prior_chase_runs <= 6), so the two groups are directly comparable and 7 serves as a buffer. This is a negative or par prediction, so a 'pass' on the positive floors is not expected. The real test is the contrast with the unexposed group and staying below the 1.12 falsification line. A reading of 1.05 or more with p < 0.05 would contradict the hypothesis."
  }

Timeline

1. Job 1816 finished

   16 Oct, 14:39 · engine:compute-1

2. H-bb6d2ec0 Discovery → KILLED fails: 33 bets, needs 300; p 0.5193 (lower tail), needs 0.05; periods [], each needs 0.99 or lower

   16 Oct, 14:39 · engine:evaluator

3. discovery result for H-bb6d2ec0 (broad rule): fail

   16 Oct, 14:39 · engine:evaluator

4. H-bb6d2ec0 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 14:39 · engine:orchestrator

5. compute-1 took job 1816 (evaluate, attempt 1)

   16 Oct, 14:39 · compute-1

6. Job 1815 finished

   16 Oct, 14:38 · engine:ai-1

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

   16 Oct, 14:38 · engine:scheduler

8. H-bb6d2ec0 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 14:38 · agent:designer@v1

9. ai-1 took job 1815 (design, attempt 1)

   16 Oct, 14:38 · ai-1

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

    16 Oct, 14:38 · engine:scheduler

11. New idea H-bb6d2ec0: Exposed young marathon handicap chasers sit at par: the lift needs an unexposed chaser (test-mechanism)

    16 Oct, 14:38 · 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.