Older horses in marathon handicap chases are over-bet (mirror of the young-stayer lift)

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-910e15e3 (Dose-response: young handicap chasers at 3m to 3m3f show a smaller lif) was predicted positive but came out the other way: A/E 0.92 on 971 bets (p 0.187 for the opposite direction). O-1a183ac0
   • H-8b0e3766 (Rival: big-field festival marathon handicaps, not youth, produce the l) passed discovery as a niche: A/E 1.23 on 63 bets (p 0.28), same-race control 0.89. O-b50a0767
   • H-98848406 (Young marathon chasers stepping up to 3m4f+ for the first time carry t) passed discovery as a niche: A/E 1.11 on 97 bets (p 0.385), same-race control 0.97. O-d3a5d866
2. AskedIs the 3m4f+ young-horse lift really the mirror of older, proven marathon chasers being over-bet in the same races? answered
3. MechanismPunters and tipsters lean on the idea that marathons need a hardened veteran. Money goes onto older, exposed horses with long staying records. By ten and over many have declining ability and stiff marks from past form, so the price overstates their chance and the young rivals drift.
4. Why the market misses itThe form book shows a long list of 3m+ placings for old horses, which looks like proven stamina. It hides that the ability behind those placings is fading and that the horse is still handicapped on it.
5. Predictednegative, A/E 0.85 to 0.95 on 500+ bets
6. Falsified if

Mechanism and rationale

Punters and tipsters lean on the idea that marathons need a hardened veteran. Money goes onto older, exposed horses with long staying records. By ten and over many have declining ability and stiff marks from past form, so the price overstates their chance and the young rivals drift.

The rule

race_type = 'Chase' AND is_handicap = 1 AND age >= 10 AND distance_yards >= 6160

Reads: race_type, is_handicap, age, distance_yards

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

2de011252e271a3ec2c1649f3ce1563fce59884e170e1a6e7e1dc2f40a1a2614

Frozen

not frozen

The definition can still change.

Confounders

• Field size and festival marathons (Grand National, Scottish/Welsh/Irish Nationals): big fields with huge overrounds depress A/E for all runners; read through the same-race control and check field_size <= 13 separately
• Top weights are disproportionately older horses: weight carried, not age, may drive the shortfall
• Going: soft-ground marathons may punish older horses differently; check by going_group
• Irish versus GB pricing and race structure; per the letters finding, a regional split is not treated as a mechanism on its own
• Price band: a shortfall carried only by 25/1+ outsiders is the general longshot bias, not veteran over-betting; the 16/1-or-shorter refinement tests this
• Period/seasonal concentration (spring festival window): check the effect across periods, since this family's young-stayer lift fades in the latest period

Control

{
  "description": "Same-race control: the other runners (mostly younger) in the same 3m4f+ handicap chases. If the mechanism holds, veterans show A/E below 1 and their same-race rivals show A/E at or above 1.0, with the age 7-and-under subset near the live system's ~1.15. If both veterans and their rivals sit below 1 (or both above), the result is a race-level property (overround, field size, festival marathons), not an age bias. A second comparison group checks that the over-bet is specific to marathons: veterans in shorter 3m to 3m3f handicap chases, which should show a smaller or no shortfall.",
  "rule": "race_type = 'Chase' AND is_handicap = 1 AND age >= 10 AND distance_yards >= 5280 AND distance_yards < 6160"
}

Discovery result

fail broad rule · 16 Oct, 13:51

Bets

148

10 winners

A/E

0.97

95%: 0.47 to 1.78

SP return

-61.3%

level stakes

Betfair SP return

-54.2%

148 bets with a BSP

Strike rate 6.8%. p = 0.58. Same-race control: the other runners in these races had A/E 1.03, 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 · 40s · 16 Oct, 13:47

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND is_handicap = 1 AND age >= 10 AND distance_yards >= 6160",
    "refinements": [
      {
        "label": "Backed veterans (16/1 or shorter)",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND age >= 10 AND distance_yards >= 6160 AND sp_decimal <= 17"
      },
      {
        "label": "Oldest band (11+)",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND age >= 11 AND distance_yards >= 6160"
      },
      {
        "label": "Exposed veterans (many prior runs)",
        "rule": "race_type = 'Chase' AND is_handicap = 1 AND age >= 10 AND distance_yards >= 6160 AND prior_runs >= 15"
      }
    ],
    "control": {
      "description": "Same-race control: the other runners (mostly younger) in the same 3m4f+ handicap chases. If the mechanism holds, veterans show A/E below 1 and their same-race rivals show A/E at or above 1.0, with the age 7-and-under subset near the live system's ~1.15. If both veterans and their rivals sit below 1 (or both above), the result is a race-level property (overround, field size, festival marathons), not an age bias. A second comparison group checks that the over-bet is specific to marathons: veterans in shorter 3m to 3m3f handicap chases, which should show a smaller or no shortfall.",
      "rule": "race_type = 'Chase' AND is_handicap = 1 AND age >= 10 AND distance_yards >= 5280 AND distance_yards < 6160"
    },
    "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": [
      "Field size and festival marathons (Grand National, Scottish/Welsh/Irish Nationals): big fields with huge overrounds depress A/E for all runners; read through the same-race control and check field_size <= 13 separately",
      "Top weights are disproportionately older horses: weight carried, not age, may drive the shortfall",
      "Going: soft-ground marathons may punish older horses differently; check by going_group",
      "Irish versus GB pricing and race structure; per the letters finding, a regional split is not treated as a mechanism on its own",
      "Price band: a shortfall carried only by 25/1+ outsiders is the general longshot bias, not veteran over-betting; the 16/1-or-shorter refinement tests this",
      "Period/seasonal concentration (spring festival window): check the effect across periods, since this family's young-stayer lift fades in the latest period"
    ],
    "mechanism": "Punters and tipsters believe marathons need a hardened veteran, so money goes onto horses aged 10+ with long 3m+ records. By ten many are declining, and their handicap marks still reflect past peak form. So their SP overstates their chance, and the younger rivals (the live young-stayer system's runners) drift to over-generous prices. If the young-stayer lift is this mirror effect, veterans in the same races should come out with A/E clearly below 1.",
    "market_miss": "The form book shows a long list of 3m+ placings for an old horse, which reads as proven stamina. It hides two things: the ability behind those placings is fading, and the horse is still handicapped on that older form. Punters price the stamina certainty and miss the decline in ability and the stiff mark.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.85 to 0.95 on 500+ bets, with same-race rivals at or above 1.0",
      "falsified_if": "A/E at or above 1.0 on 300+ bets, or a veteran shortfall that equals the same-race rivals' (both below 1). Either would mean the young lift is not the mirror of an older-horse bias."
    },
    "notes": "Direction is negative (over-bet). The vault judges it in that direction, so the A/E criterion applies inverted: A/E at or below 1/1.05 (about 0.95), p < 0.05, 300+ bets, raised from the 150 floor to match the falsification threshold. The draft rule is kept unchanged: it uses only pre-off columns and mirrors the live system's distance and handicap scope. The refinements come from the reasoning: veterans actually backed (money going on them), the oldest band (steepest decline), and heavily exposed horses (stiff marks)."
  }

Timeline

1. Job 1690 finished

   16 Oct, 13:51 · engine:compute-1

2. H-9cf148a1 Discovery → KILLED fails: 148 bets, needs 300; A/E 0.969, needs 0.95 or lower; p 0.5433 (lower tail), needs 0.05; periods [0.9], each needs 0.99 or lower

   16 Oct, 13:51 · engine:evaluator

3. discovery result for H-9cf148a1 (broad rule): fail

   16 Oct, 13:51 · engine:evaluator

4. H-9cf148a1 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 13:51 · engine:orchestrator

5. compute-1 took job 1690 (evaluate, attempt 1)

   16 Oct, 13:51 · compute-1

6. Job 1688 finished

   16 Oct, 13:48 · engine:ai-1

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

   16 Oct, 13:48 · engine:scheduler

8. H-9cf148a1 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 13:48 · agent:designer@v1

9. ai-1 took job 1688 (design, attempt 1)

   16 Oct, 13:47 · ai-1

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

    16 Oct, 13:47 · engine:scheduler

11. New idea H-9cf148a1: Older horses in marathon handicap chases are over-bet (mirror of the young-stayer lift) (rival-explanation)

    16 Oct, 13:47 · 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.