Rival: short-priced runners of all ages beat their SP in marathon handicap chases

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-3518afc1 (The higher-weight marathon lift holds at prices under 16/1): the price band 'up to 7/1' carries the effect (A/E 1.55) while the rest score up to 14/1 0.78. O-39cc6abc
   • H-d473ec0f (Rival: higher-weighted runners of all ages beat their price in maratho) failed the vault (contradicted): A/E 0.98 on 1378 bets. O-702b4301
   • H-3518afc1 (The higher-weight marathon lift holds at prices under 16/1) passed discovery as a niche: A/E 1.2 on 213 bets (p 0.182), same-race control 0.73. O-81995427
   • H-98848406 (Young marathon chasers stepping up to 3m4f+ for the first time carry t) met the sealed vault supportive: A/E 1.22 on 384 bets (p 0.0945, threshold 0.0025). O-bef6da89
2. AskedIn 3m4f+ handicap chases, do runners priced 7/1 or shorter beat their SP whatever their age or weight, so that the lift in the young slices is really a price-band effect? answered
3. MechanismOver 3m4f+ the race takes out horses that don't stay or jump, and the best-handicapped stayers win more often than in shorter races. Weight and youth are proxies. The real cause could be that the horses the market rates as genuine contenders are under-rated, because stamina makes the result less random than the betting assumes.
4. Why the market misses itMarathon handicaps carry a reputation as lotteries, with Nationals and big fields. Punters spread money onto longshots, and the books lengthen the fancied runners to balance the overround. That is a heavier favourite-longshot pattern than in ordinary races.

Mechanism and rationale

Over 3m4f+ the race takes out horses that don't stay or jump, and the best-handicapped stayers win more often than in shorter races. Weight and youth are proxies. The real cause could be that the horses the market rates as genuine contenders are under-rated, because stamina makes the result less random than the betting assumes.

The rule

race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND sp_decimal <= 8.0

Reads: race_type, is_handicap, distance_yards, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

0c935a848e81111f856855c96668a5bbd20bc5253299871f6dfd0c91b84ebe55

Frozen

not frozen

The definition can still change.

Confounders

• Overlap with the young-stayer system: short-priced runners aged 7 or under form part of the sample and could carry the lift. The age 8+ refinement must hold on its own
• Region mix: Irish marathon handicaps had fewer bets and a higher A/E in earlier slices, so a GB-versus-IRE split must not show the effect resting on Ireland alone
• Small fields with a tight SP market, where 8/1 or shorter covers most of the field and the selection is close to the whole race
• Festival marathon handicaps (Grand Nationals and similar) with unusual market dynamics and huge fields. The 13-or-fewer refinement checks this
• Earlier price-band findings: higher-weighted marathon runners drew their lift from up to 7/1 (1.55), so this rule partly re-measures a slice already seen. It is not independent evidence on the discovery data
• Overround variation: the books' percentage in big-field marathons may differ, which changes the A/E baseline

Control

{
  "description": "Same-race control: the runners priced above 8.0 in the same 3m4f+ handicap chases. If the mechanism is a sharper favourite-longshot bias in marathons, these longshots should score below par (A/E around 0.85 to 0.95), so they sit clearly under the fancied group. If the longshots also score at or above 1.0, the lift belongs to the race type (or to a soft overround) and is not a pricing error on the fancied runners. As a secondary check, the lab should compare with fancied runners (sp_decimal <= 8.0) in 3m to 3m3f handicap chases. They should sit near par, otherwise the effect is a general short-price bias and not specific to marathons.",
  "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND sp_decimal > 8.0"
}

Discovery result

pass broad rule (niche) · 16 Oct, 14:17

Bets

157

32 winners

A/E

1.17

95%: 0.80 to 1.66

SP return

+0.6%

level stakes

Betfair SP return

+15.4%

157 bets with a BSP

Strike rate 20.4%. p = 0.20. Same-race control: the other runners in these races had A/E 0.72, 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. 4 of 6 survived.

• partition: Older fancied runners only (age 8+): separate price from youthpass

  96 bets · A/E 1.19 (0.71 to 1.85) · control 0.90 · 16 Oct, 14:18

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27861.13

  after 2026-04-2710–

• perturbation: Price cut-off widened to 9/1 (sp <= 10.0)pass

  216 bets · A/E 1.11 (0.78 to 1.54) · control 0.69 · 16 Oct, 14:18

  Breakdown

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 · 49s · 16 Oct, 14:16

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND sp_decimal <= 8.0",
    "refinements": [
      {
        "label": "Older fancied runners only (age 8+), the test that separates price from youth",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND sp_decimal <= 8.0 AND age >= 8"
      },
      {
        "label": "Big fields (14+ runners), where the lottery reputation and spread-out money should be heaviest",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND sp_decimal <= 8.0 AND field_size >= 14"
      },
      {
        "label": "Ordinary-sized fields (13 or fewer), to check the lift is not only a festival or National effect",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND sp_decimal <= 8.0 AND field_size <= 13"
      }
    ],
    "control": {
      "description": "Same-race control: the runners priced above 8.0 in the same 3m4f+ handicap chases. If the mechanism is a sharper favourite-longshot bias in marathons, these longshots should score below par (A/E around 0.85 to 0.95), so they sit clearly under the fancied group. If the longshots also score at or above 1.0, the lift belongs to the race type (or to a soft overround) and is not a pricing error on the fancied runners. As a secondary check, the lab should compare with fancied runners (sp_decimal <= 8.0) in 3m to 3m3f handicap chases. They should sit near par, otherwise the effect is a general short-price bias and not specific to marathons.",
      "rule": "race_type = 'Chase' AND is_handicap = 1 AND distance_yards >= 6160 AND sp_decimal > 8.0"
    },
    "pass_criteria": {
      "min_bets": 250,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Overlap with the young-stayer system: short-priced runners aged 7 or under form part of the sample and could carry the lift. The age 8+ refinement must hold on its own",
      "Region mix: Irish marathon handicaps had fewer bets and a higher A/E in earlier slices, so a GB-versus-IRE split must not show the effect resting on Ireland alone",
      "Small fields with a tight SP market, where 8/1 or shorter covers most of the field and the selection is close to the whole race",
      "Festival marathon handicaps (Grand Nationals and similar) with unusual market dynamics and huge fields. The 13-or-fewer refinement checks this",
      "Earlier price-band findings: higher-weighted marathon runners drew their lift from up to 7/1 (1.55), so this rule partly re-measures a slice already seen. It is not independent evidence on the discovery data",
      "Overround variation: the books' percentage in big-field marathons may differ, which changes the A/E baseline"
    ],
    "mechanism": "Over 3m4f+ in handicap chases, attrition from failing to stay and from jumping errors removes much of the field. The result then depends more on genuine stamina and class than in shorter races. The runners the market already marks as real contenders (8/1 or shorter) should convert that into wins more often than their SP implies. Youth and higher weight, which showed lifts before, are proxies for 'genuine contender' and not causes.",
    "market_miss": "Punters see marathon handicaps as lotteries, especially big-field Nationals, and spread their money across longshots. To balance the overround, the books lengthen the fancied runners. That gives a steeper favourite-longshot bias than in ordinary handicaps, which leaves runners at 8/1 or shorter under-priced and longshots over-bet.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.08 to 1.20 on 250+ bets, with same-race longshots (>8.0) below par",
      "falsified_if": "Falsified if the broad rule scores A/E at or below 1.0 on 200+ bets, or if 8+ year olds at 8.0 or shorter score clearly below 7-and-under runners at the same prices, which would put the lift back on youth. It also fails if same-race longshots score at or above the fancied group."
    },
    "notes": "I kept the rule sketch: every column it uses is pre-off. sp_decimal <= 8.0 is the 7/1-or-shorter cut-off. min_bets is raised to 250 to match the prediction, and because the family's findings show that discovery samples of about 300 passed falsification and then failed in the vault. The age 8+ refinement is the decisive one for this rival explanation. Previous slices drew their lift from up to 7/1, so treat discovery overlap with caution and weight the vault and forward results."
  }

Timeline

1. Job 1795 finished

   16 Oct, 14:18 · engine:compute-1

2. H-8810c1c1 Falsification → KILLED falsified: driven by one price band (up to 7/1): the rest score 0.95; fewer than two periods with enough bets to compare (thin: the forward record will judge); IRE: too few bets to judge

   16 Oct, 14:18 · engine:falsification

3. falsification result for H-8810c1c1 (price_band: Shortest prices (3/1 or shorter) should carry the steepest favourite bias): info

   16 Oct, 14:18 · engine:evaluator

4. compute-1 took job 1795 (attack, attempt 1)

   16 Oct, 14:18 · compute-1

5. Job 1794 finished

   16 Oct, 14:18 · engine:compute-1

6. falsification result for H-8810c1c1 (partition: Ordinary fields (13 or fewer): not a festival or National effect): pass

   16 Oct, 14:18 · engine:evaluator

7. compute-1 took job 1794 (attack, attempt 1)

   16 Oct, 14:18 · compute-1

8. Job 1793 finished

   16 Oct, 14:18 · engine:compute-1

9. falsification result for H-8810c1c1 (control: Control: fancied runners in 3m to 3m3f handicap chases): fail

   16 Oct, 14:18 · engine:evaluator

10. compute-1 took job 1793 (attack, attempt 1)

    16 Oct, 14:18 · compute-1

11. Job 1792 finished

    16 Oct, 14:18 · engine:compute-1

12. falsification result for H-8810c1c1 (perturbation: Distance cut-off lowered about 10% to 3m2f+ (5600 yards)): pass

    16 Oct, 14:18 · engine:evaluator

13. compute-1 took job 1792 (attack, attempt 1)

    16 Oct, 14:18 · compute-1

14. Job 1791 finished

    16 Oct, 14:18 · engine:compute-1

15. falsification result for H-8810c1c1 (perturbation: Price cut-off widened to 9/1 (sp <= 10.0)): pass

    16 Oct, 14:18 · engine:evaluator

16. compute-1 took job 1791 (attack, attempt 1)

    16 Oct, 14:18 · compute-1

17. Job 1790 finished

    16 Oct, 14:18 · engine:compute-1

18. falsification result for H-8810c1c1 (partition: Older fancied runners only (age 8+): separate price from youth): pass

    16 Oct, 14:18 · engine:evaluator