Second-tier fancied spring handicap returners (3/1 to 7/1) beat their prices at any age

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-71a94a07 (Rival: short-priced spring handicap returners pay at any age, so H126') passed discovery as a niche: A/E 1.1 on 141 bets (p 0.271), same-race control 0.95. O-21cb941f
   • H-71a94a07 (Rival: short-priced spring handicap returners pay at any age, so H126'): the period '2025-10-01 to 2026-04-27' carries the effect (A/E 1.18) while the rest score after 2026-04-27 0.97. O-7ebaf3d7
2. AskedWithin spring Flat handicap returners after 150+ days, does the lift sit in the second tier of the market (roughly 3/1 to 7/1) rather than at 3/1 or shorter, at any age? testing
3. MechanismSome money for a returner shows the yard thinks it is ready. When a returner is clear favourite at short odds, that confidence is plain and heavily scrutinised, so the rust discount is mostly removed. A returner at 4/1 to 7/1 has partial support, so the standard 'needs the run' discount, plus a mark that has not caught up with winter progress, still holds the price up. This is a lead from one neighbour (the 4/1 cap) and is stated as such before testing the band on its own.
4. Why the market misses itThe form book shows only a long absence and an old handicap mark. Compilers price the absence as a uniform penalty. The odds shorten fully only for the obvious stable favourite, so moderately backed returners stay slightly too big.
5. Predictedpositive, A/E 1.10 to 1.25 on 250+ bets, with ages other than four at or above about 1.05
6. Falsified ifA/E below 1.03 on 250+ bets; or ages other than four below 1.0 while four-year-olds carry it; or either period below 1.0

Mechanism and rationale

Some money for a returner shows the yard thinks it is ready. When a returner is clear favourite at short odds, that confidence is plain and heavily scrutinised, so the rust discount is mostly removed. A returner at 4/1 to 7/1 has partial support, so the standard 'needs the run' discount, plus a mark that has not caught up with winter progress, still holds the price up. This is a lead from one neighbour (the 4/1 cap) and is stated as such before testing the band on its own.

The rule

race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND days_since_last_run >= 150 AND sp_decimal > 4 AND sp_decimal <= 8

Reads: race_type, is_handicap, month, days_since_last_run, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

f1c03087cbe1bfad7b4d1a711211101eb956ecd6665301e408db63914b3d08c1

Frozen

not frozen

The definition can still change.

Confounders

• The 3/1-7/1 band was picked from a perturbation neighbour (the 4/1 cap, 4/1-7/1 at 1.42 on 132) on essentially one spring, so it may be selection noise; period stability is the main check
• Overlap with the four-year-old H126 returners (1.29 on 322) could carry the whole band; the age <> 4 refinement must hold on its own
• A general mild returner lift (1.07 on 971) could explain it without any price-tier mechanism; the short-price control must score clearly lower
• A few big yards' spring campaigns could dominate one season; check trainer concentration
• Edge previously faded after late April (1.18 to 0.97), so a March-April concentration would narrow the claim
• Irish sample is tiny and Irish splits have been unreadable

Control

{
  "description": "Same-race other runners should sit near 0.95-1.00 (normal overround drag); if they also score above 1.05 the lift is a race property, not a returner effect. The key comparison is the question's own: the same spring handicap returners at 3/1 or shorter (sp_decimal <= 4). The mechanism predicts those are fully priced (A/E about 0.95-1.02), clearly below the 3/1-7/1 band. If short-priced returners score as high as or higher than the band, the price-tier mechanism fails and the result is only the general mild returner lift.",
  "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND days_since_last_run >= 150 AND sp_decimal <= 4"
}

Discovery result

pass broad rule (niche) · 16 Oct, 13:34

Bets

473

72 winners

A/E

1.08

95%: 0.85 to 1.36

SP return

-13.6%

level stakes

Betfair SP return

-1.9%

473 bets with a BSP

Strike rate 15.2%. p = 0.26. Same-race control: the other runners in these races had A/E 0.99, 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. 5 of 6 survived.

• partition: Ages other than four (separate from H126 four-year-old thread)pass

  315 bets · A/E 1.14 (0.84 to 1.50) · control 0.97 · 16 Oct, 13:35

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-272141.17

  after 2026-04-271011.06

• perturbation: Shifted band 7/2 to 6/1 (inner perturbation)pass

  278 bets · A/E 1.00 (0.71 to 1.37) · control 1.00 · 16 Oct, 13:35

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 · 32s · 16 Oct, 13:30

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND days_since_last_run >= 150 AND sp_decimal > 4 AND sp_decimal <= 8",
    "refinements": [
      {
        "label": "ages other than four (tests 'at any age', separates from the H126 four-year-old thread)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND days_since_last_run >= 150 AND sp_decimal > 4 AND sp_decimal <= 8 AND age <> 4"
      },
      {
        "label": "GB only (Irish sample tiny and Irish splits have repeatedly killed this thread)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND days_since_last_run >= 150 AND sp_decimal > 4 AND sp_decimal <= 8 AND region = 'GB'"
      }
    ],
    "control": {
      "description": "Same-race other runners should sit near 0.95-1.00 (normal overround drag); if they also score above 1.05 the lift is a race property, not a returner effect. The key comparison is the question's own: the same spring handicap returners at 3/1 or shorter (sp_decimal <= 4). The mechanism predicts those are fully priced (A/E about 0.95-1.02), clearly below the 3/1-7/1 band. If short-priced returners score as high as or higher than the band, the price-tier mechanism fails and the result is only the general mild returner lift.",
      "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND days_since_last_run >= 150 AND sp_decimal <= 4"
    },
    "pass_criteria": {
      "min_bets": 250,
      "min_ae": 1.08,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 120,
      "min_ae": 1.03,
      "max_p": 0.2
    },
    "confounders": [
      "The 3/1-7/1 band was picked from a perturbation neighbour (the 4/1 cap, 4/1-7/1 at 1.42 on 132) on essentially one spring, so it may be selection noise; period stability is the main check",
      "Overlap with the four-year-old H126 returners (1.29 on 322) could carry the whole band; the age <> 4 refinement must hold on its own",
      "A general mild returner lift (1.07 on 971) could explain it without any price-tier mechanism; the short-price control must score clearly lower",
      "A few big yards' spring campaigns could dominate one season; check trainer concentration",
      "Edge previously faded after late April (1.18 to 0.97), so a March-April concentration would narrow the claim",
      "Irish sample is tiny and Irish splits have been unreadable"
    ],
    "mechanism": "When a returner has some money behind it, that suggests the yard thinks it is ready. If the returner is a short-priced favourite (3/1 or shorter), the confidence is obvious and closely watched, so the market removes most of the rust discount. If it has only partial support (just over 3/1 up to 7/1), the usual 'needs the run' discount still applies, along with a handicap mark that has not caught up with winter progress, so the price stays too big. This lead comes from one neighbouring result (the 4/1 cap) and is stated as such before the band is tested on its own, with no age restriction.",
    "market_miss": "All the form book shows is a long absence and an old handicap mark. Compilers treat the absence as a flat penalty, and the price shortens fully only for the obvious stable favourite. Returners with moderate support, at just over 3/1 to 7/1, stay slightly too big because the market reads partial support as weak evidence of readiness.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.10 to 1.25 on 250+ bets; ages other than four at or above about 1.05; short-priced (<=3/1) returner control near 1.0",
      "falsified_if": "A/E below 1.03 on 250+ bets; or ages other than four below 1.0 while four-year-olds carry it; or either period below 1.0; or the <=3/1 returner control scoring at or above the band"
    },
    "notes": "The rule sketch is kept as written. It uses only pre-off columns (race_type, is_handicap, month, days_since_last_run, sp_decimal) and has no age or prior_runs limit, as the 'any age' question needs. 'sp_decimal > 4' leaves out exactly 3/1, so the band starts just above 3/1 and runs to 7/1 inclusive; this also keeps it apart from the control, so no runner can fall in both. Pass criteria are stricter than the floors (250 bets, 1.08) because the prediction itself names 250+ bets and a 1.10+ size. Refinements are looked at only if the broad rule shows evidence."
  }

Timeline

1. Job 1642 finished

   16 Oct, 13:35 · engine:compute-1

2. H-12ef9cb4 Falsification → KILLED falsified: IRE A/E 0.879

   16 Oct, 13:35 · engine:falsification

3. falsification result for H-12ef9cb4 (period: March-April only partition): pass

   16 Oct, 13:35 · engine:evaluator

4. compute-1 took job 1642 (attack, attempt 1)

   16 Oct, 13:35 · compute-1

5. Job 1641 finished

   16 Oct, 13:35 · engine:compute-1

6. falsification result for H-12ef9cb4 (control: Control: same price band, recent runners (not returners)): fail

   16 Oct, 13:35 · engine:evaluator

7. compute-1 took job 1641 (attack, attempt 1)

   16 Oct, 13:35 · compute-1

8. Job 1640 finished

   16 Oct, 13:35 · engine:compute-1

9. falsification result for H-12ef9cb4 (perturbation: Absence threshold perturbed to 120+ and 180+ days): pass

   16 Oct, 13:35 · engine:evaluator

10. compute-1 took job 1640 (attack, attempt 1)

    16 Oct, 13:35 · compute-1

11. Job 1639 finished

    16 Oct, 13:35 · engine:compute-1

12. falsification result for H-12ef9cb4 (perturbation: Widened band 5/2 to 9/1 (outer perturbation)): pass

    16 Oct, 13:35 · engine:evaluator

13. compute-1 took job 1639 (attack, attempt 1)

    16 Oct, 13:35 · compute-1

14. Job 1638 finished

    16 Oct, 13:35 · engine:compute-1

15. falsification result for H-12ef9cb4 (perturbation: Shifted band 7/2 to 6/1 (inner perturbation)): pass

    16 Oct, 13:35 · engine:evaluator

16. compute-1 took job 1638 (attack, attempt 1)

    16 Oct, 13:35 · compute-1

17. Job 1637 finished

    16 Oct, 13:35 · engine:compute-1

18. falsification result for H-12ef9cb4 (partition: Ages other than four (separate from H126 four-year-old thread)): pass

    16 Oct, 13:35 · engine:evaluator

19. compute-1 took job 1637 (attack, attempt 1)

    16 Oct, 13:35 · compute-1

20. Job 1636 finished

    16 Oct, 13:35 · engine:ai-1