Lightly raced five-year-olds back from a break: is age the cause?

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
   • H126-lightly-raced-four-year-old-returns (H126: lightly raced four year old returns) was GREEN in the earlier research: A/E 1.12 on 2958 bets (periods 1.21 / 1.01 / 1.20); now PROMISING. O-7b05ae3b
2. AskedIs the H126 lift caused by four-year-old physical maturation that the handicap mark hasn't caught up with, or just by lightly raced horses returning fresh at any age? answered
3. MechanismIf the cause is growing stronger between three and four, five-year-olds with the same light record and long break should show little or no lift. If they lift as much, the cause is freshness or a light record, not age.
4. Why the market misses itUnder the maturation story, the market reads the old three-year-old form and mark; a five-year-old has had another season for the handicapper to catch up.
5. Predictednegative, A/E 0.95 to 1.03 on 400+ bets (no edge), well below H126's 1.12
6. Falsified ifFive-year-olds score A/E 1.10 or more on 400+ bets with every period above 1.0. That would mean light racing plus a break, not age, carries the edge.
7. Rule sketchrace_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 5 AND prior_runs <= 10 AND days_since_last_run >= 150
8. question idQ-dc3ec26a

The question's answer so far: Not maturation. In the vault the four-year-old March-April rule scored 1.02 on 1548 and all spring returners 1.03 on 6934. Non-four-year-olds scored 1.048 on 965 at discovery. Age four adds nothing measurable once tested out of sample.

Mechanism and rationale

If the cause is growing stronger between three and four, five-year-olds with the same light record and long break should show little or no lift. If they lift as much, the cause is freshness or a light record, not age.

The rule

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

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

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

a4cdfe46a26f85e9eee8b0512f9c1e22155ad5d2ee2ad3d8cda1a7024222a4f3

Frozen

not frozen

The definition can still change.

Confounders

• Five-year-olds with 10 or fewer runs since July 2020 are a different population: injury-prone or late-developing horses, horses switched from jumps or from abroad, or horses whose earlier career falls before the prior_runs window. Any of these can raise or lower the lift for reasons that have nothing to do with age.
• prior_runs only counts from July 2020, so in 2020-22 some five-year-olds look lightly raced when they are not. Split by period, and give the latest periods more weight.
• Smaller sample than H126: some 400 or more bets are needed before a 'no lift' reading means anything rather than just noise.
• Different trainer mix: yards that keep older horses lightly raced (patient big yards, or dual-purpose jumps yards) may carry their own effect.
• Longshot artefact: a lift that only shows at 16/1+ comes from the yardstick, not the mechanism, so read the result in the 14/1-or-shorter band.
• Class and mark: five-year-olds have had longer to be assessed, so they may run off fairer marks or in lower classes, which mixes 'handicapper has caught up' with plain age.

Discovery result

fail broad rule · 16 Oct, 12:11

Bets

89

6 winners

A/E

0.76

95%: 0.28 to 1.66

SP return

-69.3%

level stakes

Betfair SP return

-67.5%

89 bets with a BSP

Strike rate 6.7%. p = 0.79. 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 · 34s · 16 Oct, 12:09

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 5 AND prior_runs <= 10 AND days_since_last_run >= 150",
    "refinements": [
      {
        "label": "14/1 or shorter (strip the known longshot artefact)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 5 AND prior_runs <= 10 AND days_since_last_run >= 150 AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "The real comparison is H126: the identical rule with age = 4 (A/E 1.12 on 2958). Everything except age stays the same, so the gap between the two A/Es measures the age effect alone. The lab's usual same-race comparison (other runners in these races) should sit near 1.00. If the five-year-olds also sit near 1.00 while the four-year-olds sit at about 1.12, maturation is the cause. If five-year-olds match the four-year-olds, freshness plus a light record is the cause and age is not.",
      "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"
    },
    "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": [
      "Five-year-olds with 10 or fewer runs since July 2020 are a different population: injury-prone or late-developing horses, horses switched from jumps or from abroad, or horses whose earlier career falls before the prior_runs window. Any of these can raise or lower the lift for reasons that have nothing to do with age.",
      "prior_runs only counts from July 2020, so in 2020-22 some five-year-olds look lightly raced when they are not. Split by period, and give the latest periods more weight.",
      "Smaller sample than H126: some 400 or more bets are needed before a 'no lift' reading means anything rather than just noise.",
      "Different trainer mix: yards that keep older horses lightly raced (patient big yards, or dual-purpose jumps yards) may carry their own effect.",
      "Longshot artefact: a lift that only shows at 16/1+ comes from the yardstick, not the mechanism, so read the result in the 14/1-or-shorter band.",
      "Class and mark: five-year-olds have had longer to be assessed, so they may run off fairer marks or in lower classes, which mixes 'handicapper has caught up' with plain age."
    ],
    "mechanism": "This tests a rival explanation for H126. If the four-year-old lift comes from physical maturation between three and four, which the old form and mark do not yet reflect, then five-year-olds with the same light record and a 150+ day break should show little or no lift, because their growth is largely done and is already in their form. If five-year-olds lift as much as the four-year-olds, the cause is freshness or a light record, and age is not.",
    "market_miss": "Under the maturation story, the market and the handicapper price a returning four-year-old off three-year-old form that understates its current strength. A five-year-old has had another season for its form and mark to catch up, so the market should price it correctly and there should be nothing to miss. A lift here would mean the market misprices freshness or light racing in general, not age.",
    "prediction": {
      "direction": "negative",
      "size": "A/E about 0.95 to 1.03 on 400+ bets: no edge, well below H126's 1.12 under the same rule with age = 4",
      "falsified_if": "Five-year-olds score A/E 1.10 or more on 400+ bets with every period (2020-22, 2023 to a year ago, the last year) above 1.0, and the lift holds at 14/1 or shorter. That would mean light racing plus a break carries the edge, not age."
    },
    "notes": "This is a rival-explanation test, so the expected result is a null (no edge). The direction is pre-registered as 'negative' (does not beat its prices), as the hypothesis predicts. The rule is the H126 rule with only age changed (4 to 5), so the comparison isolates one variable. All the columns are pre-off and already used by H126. Read the result against the H126 control rule, not only against 1.00. A thin result of 1.03 to 1.08 is ambiguous and should go to falsification rather than being taken as support for either story. The single refinement removes 16/1+ runners because of the longshot artefact recorded in the findings. No other slicing is justified by the reasoning."
  }

Timeline

1. Job 1259 finished

   16 Oct, 12:11 · engine:compute-1

2. H-ef8fb404 Discovery → KILLED fails: 89 bets, needs 300; p 0.332 (lower tail), needs 0.05; periods [1.1], each needs 0.99 or lower

   16 Oct, 12:11 · engine:evaluator

3. discovery result for H-ef8fb404 (broad rule): fail

   16 Oct, 12:11 · engine:evaluator

4. H-ef8fb404 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 12:11 · engine:orchestrator

5. compute-1 took job 1259 (evaluate, attempt 1)

   16 Oct, 12:11 · compute-1

6. Job 1256 finished

   16 Oct, 12:10 · engine:ai-1

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

   16 Oct, 12:10 · engine:scheduler

8. H-ef8fb404 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 12:10 · agent:designer@v1

9. ai-1 took job 1256 (design, attempt 1)

   16 Oct, 12:09 · ai-1

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

    16 Oct, 12:09 · engine:scheduler

11. New idea H-ef8fb404: Lightly raced five-year-olds back from a break: is age the cause? (rival-explanation)

    16 Oct, 12:09 · 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.