Lightly raced three-year-olds on their first run after the winter in handicaps

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-ef8fb404 (Lightly raced five-year-olds back from a break: is age the cause?) pointed the predicted way but failed the bar: A/E 0.76 on 89 bets (p 0.795), control 1.03. O-cee491b3
   • H-625f1f2d (Fewest prior runs: the mark has the least information) passed discovery as a niche: A/E 1.18 on 218 bets (p 0.236), same-race control 0.98. O-e2338df0
2. AskedDo lightly raced three-year-olds on their first handicap run after the winter show the same lift, given their marks rest on even thinner juvenile form? answered
3. MechanismThree-year-olds grow and strengthen most over their first winter. A handicap mark set on a few two-year-old runs lags that growth even more than a three-year-old mark lags a four-year-old.
4. Why the market misses itThe form shows a few modest juvenile runs and a long absence. Punters and the market anchor on that form, while trainers know how much the horse has strengthened.
5. Predictedpositive, A/E 1.08 to 1.15 on 600+ bets
6. Falsified ifA/E below 1.03 on 600+ bets, or any period below 1.0
7. Rule sketchrace_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 6 AND days_since_last_run >= 150
8. question idQ-31d115c9

The question's answer so far: No. Lightly raced three-year-olds back from the winter scored A/E 1.04 on 568 bets (16/1 or bigger at 0.58) and failed the bar, so the effect does not extend one year younger.

Mechanism and rationale

Three-year-olds grow and strengthen most over their first winter. A handicap mark set on a few two-year-old runs lags that growth even more than a three-year-old mark lags a four-year-old.

The rule

race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 6 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)

91c13cedd7fb87be1bd224ad31812fd872be8321ebac21874112f74d40a6bbbb

Frozen

not frozen

The definition can still change.

Confounders

• The three-year-old handicap market is already known to favour unexposed types (H111 scored 1.02), so the lift has to beat the recently raced three-year-old control, not just 1.0
• Big-yard concentration: a few large yards send out well-handled backward juveniles, so check that no single trainer supplies most of the A/E
• Debut-season nursery marks may already be generous, which would make this a nursery-mark effect and not a maturation effect
• Surface and period confound: March and early April are mostly all-weather and May is turf, so any period concentration needs a surface check before it is read as instability
• Longshot artefact: lifts at 16/1 or bigger are common across rules and are not evidence for this trigger
• SP versus BSP: the sister H126 edge only survived at BSP, so a lift measured at SP may not be bettable

Control

{
  "description": "Same-race other runners should score about 0.95 to 1.00, so the lift is not a property of spring three-year-old handicaps as a whole. A second comparison group: lightly raced three-year-olds in the same spring handicaps that have run within the last 60 days. They have done the same winter growing, but their mark has since been updated on recent runs. If the mechanism is a mark that lags maturation, this group should score about 1.00 and the fresh group should beat it by at least 0.05. If both score the same, the edge belongs to unexposed spring three-year-olds in general and not to the lag in the mark after a break.",
  "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 6 AND days_since_last_run <= 60"
}

Discovery result

fail broad rule · 16 Oct, 12:20

Bets

568

54 winners

A/E

1.04

95%: 0.78 to 1.36

SP return

-31.0%

level stakes

Betfair SP return

-17.6%

568 bets with a BSP

Strike rate 9.5%. p = 0.41. Same-race control: the other runners in these races had A/E 1.00, 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 · 35s · 16 Oct, 12:16

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 6 AND days_since_last_run >= 150",
    "refinements": [
      {
        "label": "Least exposed: 4 or fewer prior runs",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 4 AND days_since_last_run >= 150"
      },
      {
        "label": "14/1 or shorter (exclude the known longshot artefact)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 6 AND days_since_last_run >= 150 AND sp_decimal <= 15"
      },
      {
        "label": "Turf only (separate the early-spring all-weather period from the turf season)",
        "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 6 AND days_since_last_run >= 150 AND surface = 'Turf'"
      }
    ],
    "control": {
      "description": "Same-race other runners should score about 0.95 to 1.00, so the lift is not a property of spring three-year-old handicaps as a whole. A second comparison group: lightly raced three-year-olds in the same spring handicaps that have run within the last 60 days. They have done the same winter growing, but their mark has since been updated on recent runs. If the mechanism is a mark that lags maturation, this group should score about 1.00 and the fresh group should beat it by at least 0.05. If both score the same, the edge belongs to unexposed spring three-year-olds in general and not to the lag in the mark after a break.",
      "rule": "race_type = 'Flat' AND is_handicap = 1 AND month IN (3, 4, 5) AND age = 3 AND prior_runs <= 6 AND days_since_last_run <= 60"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 120,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "The three-year-old handicap market is already known to favour unexposed types (H111 scored 1.02), so the lift has to beat the recently raced three-year-old control, not just 1.0",
      "Big-yard concentration: a few large yards send out well-handled backward juveniles, so check that no single trainer supplies most of the A/E",
      "Debut-season nursery marks may already be generous, which would make this a nursery-mark effect and not a maturation effect",
      "Surface and period confound: March and early April are mostly all-weather and May is turf, so any period concentration needs a surface check before it is read as instability",
      "Longshot artefact: lifts at 16/1 or bigger are common across rules and are not evidence for this trigger",
      "SP versus BSP: the sister H126 edge only survived at BSP, so a lift measured at SP may not be bettable"
    ],
    "mechanism": "Three-year-olds do most of their physical growing and strengthening over their first winter. A handicap mark set on 3 to 6 two-year-old runs, last updated 150+ days ago, does not reflect that growth. The gap should be at least as large as the one already seen for lightly raced four-year-olds (H126, A/E 1.12), because the mark rests on juvenile form and the horse gains more in its first winter.",
    "market_miss": "The form only shows a few modest juvenile runs followed by a long absence, and the market prices on that visible form and the unchanged mark. Trainers have seen at home how much the horse has strengthened, but nothing in the public form shows it until it runs.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.08 to 1.15 on 600+ bets",
      "falsified_if": "A/E below 1.03 on 600+ bets; or any period with 150+ bets below 1.0 after the surface check; or no lift of at least 0.05 over the recently raced (60 days or less) three-year-old control"
    },
    "notes": "I kept the rule sketch unchanged; every column it uses is pre-off and was used in earlier rules from this family. Runners with no previous run have no days_since_last_run, so the rule leaves them out. The refinements mirror what the reasoning and earlier findings point to: the 5-or-fewer-runs cut that lifted H-625f1f2d, the 14/1-or-shorter band, and the surface split. min_bets is raised to 300 because the prediction expects 600+ bets."
  }

Timeline

1. Job 1293 finished

   16 Oct, 12:20 · engine:compute-1

2. H-3449e2e4 Discovery → KILLED fails: A/E 1.04, needs 1.05; p 0.4051, needs 0.05 or lower (not a niche pass: A/E 1.04 below 1.05)

   16 Oct, 12:20 · engine:evaluator

3. discovery result for H-3449e2e4 (broad rule): fail

   16 Oct, 12:20 · engine:evaluator

4. H-3449e2e4 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 12:20 · engine:orchestrator

5. compute-1 took job 1293 (evaluate, attempt 1)

   16 Oct, 12:20 · compute-1

6. Job 1290 finished

   16 Oct, 12:16 · engine:ai-1

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

   16 Oct, 12:16 · engine:scheduler

8. H-3449e2e4 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 12:16 · agent:designer@v1

9. ai-1 took job 1290 (design, attempt 1)

   16 Oct, 12:16 · ai-1

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

    16 Oct, 12:13 · engine:scheduler

11. New idea H-3449e2e4: Lightly raced three-year-olds on their first run after the winter in handicaps (generalise)

    16 Oct, 12:13 · 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.