Irish lightly raced four-year-old returners in spring handicaps are fully priced

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-ee2c7a96 (Exposed four-year-olds back from the winter in spring handicaps: is li) pointed the predicted way but failed the bar: A/E 0.8 on 191 bets (p 0.814), control 1.02. O-2b083477
   • H-1a3e129f (Lightly raced four-year-old returners in Flat non-handicaps: progress ): the country 'GB' carries the effect (A/E 1.21) while the rest score IRE 0.86. O-b0af52cb
2. AskedIs the four-year-old spring returner edge a British phenomenon: do Irish runners in the H126 base score at or below their prices while British runners carry the whole lift? answered
3. MechanismIrish Flat racing is concentrated in a few big yards. Their lightly raced four-year-olds are closely followed, and their winter progress is well signalled by stable form and market moves. British returners come from a wider spread of yards, where winter improvement is less visible to punters.
4. Why the market misses itIn Britain the form book shows only three-year-old runs and a long absence, so the market leans on old form. In Ireland the same information is priced in through stable reputation and a well-informed market, so there is nothing left to miss.
5. Predictednegative, A/E 0.80 to 1.00 on 300+ Irish bets, against about 1.15 for GB in the same base
6. Falsified ifIrish runners score A/E 1.08 or more on 300+ bets, or match the GB figure within 0.05

Mechanism and rationale

Irish Flat racing is concentrated in a few big yards. Their lightly raced four-year-olds are closely followed, and their winter progress is well signalled by stable form and market moves. British returners come from a wider spread of yards, where winter improvement is less visible to punters.

The rule

race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND region = 'IRE'

Reads: race_type, is_handicap, month, age, prior_runs, days_since_last_run, region

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

aee6e10ba2481d690e3aa87d597f6908d28247563d328e52c96ac3451cafc502

Frozen

not frozen

The definition can still change.

Confounders

• Small Irish sample: H126 had 2958 bets in total, but the Irish share could fall well short of the 300 the prediction needs, which would leave the CI too wide to separate 0.90 from 1.10
• Irish spring handicaps have bigger fields and more longshots. Earlier Irish blanks were at 16/1+ (0 from 14 on the AW, 0 from 21 among exposed returners), so a shortfall could just be the longshot bias. The 14/1 and field-size refinements test this
• A few dominant Irish yards (e.g. Ballydoyle and Weld types) could drive the result either way. Check the trainer concentration among the bets
• Irish spring turf starts later and Ireland has a single AW track (Dundalk), so the surface mix and month mix differ from GB. That makes it a different race population, not a pricing difference
• Irish data coverage of some columns starts later (race_class only from 2026-07). The rule avoids such columns, but compare the regions within the same period
• Multiple testing: earlier Irish cuts in this family already pointed negative, so this result is partly a confirmation of observations that motivated it (O-b0af52cb, O-2b083477)

Discovery result

fail broad rule · 16 Oct, 12:43

Bets

135

10 winners

A/E

1.35

95%: 0.65 to 2.49

SP return

-18.9%

level stakes

Betfair SP return

+3.5%

135 bets with a BSP

Strike rate 7.4%. p = 0.21. Same-race control: the other runners in these races had A/E 0.96, so the selection 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 · 33s · 16 Oct, 12:36

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND region = 'IRE'",
    "refinements": [
      {
        "label": "14/1 or shorter: is the Irish shortfall more than a longshot drag?",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND region = 'IRE' AND sp_decimal <= 15"
      },
      {
        "label": "Fields of 14 or fewer: does the shortfall survive outside big Irish handicap fields?",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND region = 'IRE' AND field_size <= 14"
      }
    ],
    "control": {
      "description": "Two comparisons. (1) Same-race control: the other runners in the same Irish spring handicaps should score about 1.00, so any Irish returner shortfall is not a property of the races. (2) The key contrast is the British half of the same base (H126 restricted to GB), which is expected to score about 1.15. The hypothesis is supported only if the Irish returners sit at or below 1.00 AND at least 0.10 below the GB returners. If both regions sit near 1.10, the H126 lift is not confined to Britain and the hypothesis fails.",
      "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 4 AND prior_runs <= 10 AND days_since_last_run >= 150 AND region = 'GB'"
    },
    "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": [
      "Small Irish sample: H126 had 2958 bets in total, but the Irish share could fall well short of the 300 the prediction needs, which would leave the CI too wide to separate 0.90 from 1.10",
      "Irish spring handicaps have bigger fields and more longshots. Earlier Irish blanks were at 16/1+ (0 from 14 on the AW, 0 from 21 among exposed returners), so a shortfall could just be the longshot bias. The 14/1 and field-size refinements test this",
      "A few dominant Irish yards (e.g. Ballydoyle and Weld types) could drive the result either way. Check the trainer concentration among the bets",
      "Irish spring turf starts later and Ireland has a single AW track (Dundalk), so the surface mix and month mix differ from GB. That makes it a different race population, not a pricing difference",
      "Irish data coverage of some columns starts later (race_class only from 2026-07). The rule avoids such columns, but compare the regions within the same period",
      "Multiple testing: earlier Irish cuts in this family already pointed negative, so this result is partly a confirmation of observations that motivated it (O-b0af52cb, O-2b083477)"
    ],
    "mechanism": "Irish Flat racing is concentrated in a handful of big yards whose lightly raced four-year-olds are closely watched. Their winter progress is signalled through stable form, gallops reports and market moves, so a returner's price already reflects it. British returners come from many more yards, where winter improvement is invisible and the market leans on the three-year-old form. That is where the H126 lift (about 1.10 to 1.15) comes from.",
    "market_miss": "In Britain the form book shows only three-year-old runs and a 150+ day absence, so prices anchor to old form and understate winter progress. In Ireland the same information reaches the market through stable reputation and an informed betting ring, so nothing is left to miss. Fashionable-yard runners may even be over-bet, which would push A/E below 1.",
    "prediction": {
      "direction": "negative",
      "size": "Irish returners A/E 0.80 to 1.00 on 300+ bets, at least 0.10 below the GB half of the same base (about 1.15)",
      "falsified_if": "Irish returners score A/E 1.08 or more on 300+ bets, or come within 0.05 of the GB returners' A/E in the same base"
    },
    "notes": "The question is rival-explanation: whether the H126 four-year-old returner lift is British-only. The pre-registered direction is negative, meaning Irish returners are fully priced or over-bet. The pass criteria use the lab floors, read in the negative direction: the strength of the shortfall must reach the floor-equivalent of A/E 1.05, i.e. A/E of about 0.95 or below, with p 0.05 on 150+ bets. If the Irish sample falls below 150 bets, report it as thin, not as a failure. The GB contrast rule is the decisive comparison and must be run on the same dates. The refinements look at whether any shortfall is only the Irish longshot and big-field effect, rather than efficient pricing of credible runners."
  }

Timeline

1. Job 1401 finished

   16 Oct, 12:43 · engine:compute-1

2. H-52d3c28e Discovery → KILLED fails: 135 bets, needs 300; A/E 1.353, needs 0.95 or lower; p 0.8712 (lower tail), needs 0.05; periods [1.08], each needs 0.99 or lower; A/E 1.353 against same-race control 0.962, needs 0.03 below

   16 Oct, 12:43 · engine:evaluator

3. discovery result for H-52d3c28e (broad rule): fail

   16 Oct, 12:43 · engine:evaluator

4. H-52d3c28e Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 12:43 · engine:orchestrator

5. compute-1 took job 1401 (evaluate, attempt 1)

   16 Oct, 12:43 · compute-1

6. Job 1400 finished

   16 Oct, 12:37 · engine:ai-1

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

   16 Oct, 12:37 · engine:scheduler

8. H-52d3c28e Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 12:37 · agent:designer@v1

9. ai-1 took job 1400 (design, attempt 1)

   16 Oct, 12:36 · ai-1

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

    16 Oct, 12:36 · engine:scheduler

11. New idea H-52d3c28e: Irish lightly raced four-year-old returners in spring handicaps are fully priced (rival-explanation)

    16 Oct, 12:36 · 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.