Older higher-weighted marathon chasers match the young: weight position, not youth, carries the system

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-d473ec0f (Rival: higher-weighted runners of all ages beat their price in maratho) passed discovery as a niche: A/E 1.21 on 338 bets (p 0.155), same-race control 0.6. O-4afbccc8
   • H-ebbea470 (Rival: light weight, not youth, carries the marathon handicap chase li) was predicted positive but came out the other way: A/E 0.48 on 124 bets (p 0.08 for the opposite direction). O-ccf039fe
2. AskedOnce weight position is held fixed, does youth add anything in marathon handicap chases? answered
3. MechanismIf class, shown by being high in the weights, is what makes a marathon handicapper beat its price, then horses aged 8+ high in the weights should beat their price by as much as young ones. Young horses look good only because most of them are well up the weights.
4. Why the market misses itThe market marks down big weights over marathon trips whatever the horse's age. The young system's lift would then be the same mispricing seen through an age filter.
5. Predictedpositive, A/E 1.10 to 1.20 on 350+ bets, within 0.05 of young higher-weighted runners
6. Falsified ifA/E at or below 1.0 on 300+ bets, which would mean youth is needed and weight position alone does not carry the lift
7. Rule sketch

Mechanism and rationale

If class, shown by being high in the weights, is what makes a marathon handicapper beat its price, then horses aged 8+ high in the weights should beat their price by as much as young ones. Young horses look good only because most of them are well up the weights.

The rule

race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND age >= 8 AND weight_lbs > bottom_weight_lbs + 7

Reads: race_type, is_handicap, distance_yards, age, weight_lbs, bottom_weight_lbs

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

b043efc29683675473d843d2f1b7aad6374c9374eda7123aba8e15623ade1a05

Frozen

16 Oct, 14:17

Nothing in the definition can change after this.

Confounders

• Longshot wins: the all-age higher-weight lift relied on 4 wins at 16/1+ (1.28 on 125, CI 0.84 to 1.68). The 16/1-or-shorter refinement checks this
• Not independent: this population largely is the 1.154-on-352 'heavier older horses' slice already seen, so discovery partly re-reads data that is already known. The forward record is the real test
• Big-field festival marathon handicaps (Grand National type races) may produce lift whatever the horse's age or weight
• Class/rating: weight above bottom stands in for official rating, so the effect may be a class effect that weight position only labels
• Survivorship among older horses: 8+ year olds still high in the weights have stayed sound and well rated, which is a selected group
• The age 8-9 and 10+ bands each sat near par (0.97). Any lift here must come entirely from the weight split inside those bands

Control

{
  "description": "Same-race other runners: these are mostly the young runners and the lighter runners within 7 lb of bottom weight. If weight position carries the lift, the lighter runners should show well under par (the earlier 0.48–0.60) and the overall same-race control should sit below the rule. The direct comparison for the question 'does youth add anything once weight is fixed' is young higher-weighted runners in the same type of race. The rule's A/E should come within 0.05 of theirs. If the young group beats the older group by more than 0.10, then youth adds something.",
  "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND age <= 7 AND weight_lbs > bottom_weight_lbs + 7"
}

Discovery result

pass broad rule (niche) · 16 Oct, 14:13

Bets

253

23 winners

A/E

1.18

95%: 0.75 to 1.77

SP return

-24.2%

level stakes

Betfair SP return

-0.7%

253 bets with a BSP

Strike rate 9.1%. p = 0.24. Same-race control: the other runners in these races had A/E 0.87, so the selection beat its own races. Zone: discovery. Evaluator eval-1.0, data 6a4e98be18e83933

Falsification

Each attack is a deliberate attempt to kill the idea. 6 of 6 survived.

• perturbation: Weight cut loosened to +5 lb above bottom weightpass

  283 bets · A/E 1.16 (0.75 to 1.71) · control 0.86 · 16 Oct, 14:17

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-272611.16

  after 2026-04-2722–

• perturbation: Weight cut tightened to +9 lb above bottom weightpass

  217 bets · A/E 1.32 (0.84 to 1.98) · control 0.80 · 16 Oct, 14:17

  Breakdown

Sealed holdout (vault)

Bets

1,129

16 Oct, 14:17

A/E

0.90

p = 0.83

BH threshold then

0.0023

over every vault test so far

Outcome

Failed

Forward record

No forward record. Shadow tracking starts after the vault and counts only races run after the freeze.

Lineage

Agent runs

• designer v1ok · 38s · 16 Oct, 14:10

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND age >= 8 AND weight_lbs > bottom_weight_lbs + 7",
    "refinements": [
      {
        "label": "16/1 or shorter (strip the longshot wins that carry the all-age lift)",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND age >= 8 AND weight_lbs > bottom_weight_lbs + 7 AND sp_decimal <= 17"
      },
      {
        "label": "Ages 8-9 only (the band that sat near par overall, testing whether weight position rescues it)",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND age >= 8 AND age <= 9 AND weight_lbs > bottom_weight_lbs + 7"
      }
    ],
    "control": {
      "description": "Same-race other runners: these are mostly the young runners and the lighter runners within 7 lb of bottom weight. If weight position carries the lift, the lighter runners should show well under par (the earlier 0.48–0.60) and the overall same-race control should sit below the rule. The direct comparison for the question 'does youth add anything once weight is fixed' is young higher-weighted runners in the same type of race. The rule's A/E should come within 0.05 of theirs. If the young group beats the older group by more than 0.10, then youth adds something.",
      "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND age <= 7 AND weight_lbs > bottom_weight_lbs + 7"
    },
    "pass_criteria": {
      "min_bets": 250,
      "min_ae": 1.08,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1.05,
      "max_p": 0.2
    },
    "confounders": [
      "Longshot wins: the all-age higher-weight lift relied on 4 wins at 16/1+ (1.28 on 125, CI 0.84 to 1.68). The 16/1-or-shorter refinement checks this",
      "Not independent: this population largely is the 1.154-on-352 'heavier older horses' slice already seen, so discovery partly re-reads data that is already known. The forward record is the real test",
      "Big-field festival marathon handicaps (Grand National type races) may produce lift whatever the horse's age or weight",
      "Class/rating: weight above bottom stands in for official rating, so the effect may be a class effect that weight position only labels",
      "Survivorship among older horses: 8+ year olds still high in the weights have stayed sound and well rated, which is a selected group",
      "The age 8-9 and 10+ bands each sat near par (0.97). Any lift here must come entirely from the weight split inside those bands"
    ],
    "mechanism": "In 3m4f+ handicap chases, being high in the weights marks class: a horse rated well enough to carry more than 7 lb above bottom weight. The hypothesis says class, not youth, is why these runners beat their price. If so, horses aged 8+ high in the weights should beat their price by about as much as young higher-weighted ones. The young system would then look good only because most young marathon runners are well up the weights.",
    "market_miss": "Over marathon trips, punters and odds compilers mark down big weights whatever the horse's age, on the folk belief that weight anchors stayers over long trips. That pushes higher-weighted runners out in price. The young-stayer system's lift would then be this same weight mispricing seen through an age filter.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.10 to 1.20 on 300+ bets, within 0.05 of young higher-weighted runners in the same races",
      "falsified_if": "A/E at or below 1.0 on 300+ bets, or more than 0.10 below the young higher-weighted control. Either would mean youth is needed and weight position alone does not carry the lift. Also falsified if the lift disappears (A/E at or below 1.0) at 16/1 or shorter."
    },
    "notes": "I kept the rule sketch unchanged. weight_lbs is used in rules already measured for this family. The pass bar is stricter than the floor (250 bets, A/E 1.08) because this population overlaps the 1.154-on-352 slice already seen, so discovery is not a clean out-of-sample test. The forward bar is A/E 1.05 so the live record has to confirm a real lift, not just par. Refinements come only from the reasoning: the longshot confound and the 8-9 band that sat near par."
  }

Timeline

1. Job 1788 finished

   16 Oct, 14:17 · engine:compute-1

2. H-b9326edc Vault → KILLEDfailed the vault vault CONTRADICTED: A/E 0.903 on 1129 bets, p 0.8329 against BH threshold 0.002326 over 43 vault tests

   16 Oct, 14:17 · engine:orchestrator

3. vault result for H-b9326edc (vault query (contradicted)): fail

   16 Oct, 14:17 · engine:orchestrator

4. H-b9326edc Frozen → FROZENvault vault query started

   16 Oct, 14:17 · engine:orchestrator

5. compute-1 took job 1788 (vault, attempt 1)

   16 Oct, 14:17 · compute-1

6. Job 1778 finished

   16 Oct, 14:17 · engine:compute-1

7. Queued vault on the compute lane: frozen: one sealed vault query

   16 Oct, 14:17 · engine:scheduler

8. H-b9326edc Falsification → FROZEN survived falsification

   16 Oct, 14:17 · engine:falsification

9. falsification result for H-b9326edc (control: Control: same older higher-weight runners at 3m to 3m3f): pass

   16 Oct, 14:17 · engine:evaluator

10. compute-1 took job 1778 (attack, attempt 1)

    16 Oct, 14:17 · compute-1

11. Job 1777 finished

    16 Oct, 14:17 · engine:compute-1

12. falsification result for H-b9326edc (partition: Exclude big-field festival marathons (field 16 or fewer)): pass

    16 Oct, 14:17 · engine:evaluator

13. compute-1 took job 1777 (attack, attempt 1)

    16 Oct, 14:17 · compute-1

14. Job 1776 finished

    16 Oct, 14:17 · engine:compute-1

15. falsification result for H-b9326edc (price_band: 16/1 or shorter only (strip the 4 longshot wins)): pass

    16 Oct, 14:17 · engine:evaluator

16. compute-1 took job 1776 (attack, attempt 1)

    16 Oct, 14:17 · compute-1

17. Job 1775 finished

    16 Oct, 14:17 · engine:compute-1

18. falsification result for H-b9326edc (perturbation: Trip cut lowered about 10% to 3m1f+): pass

    16 Oct, 14:17 · engine:evaluator

19. compute-1 took job 1775 (attack, attempt 1)

    16 Oct, 14:17 · compute-1