Negative: short-priced new-yard runners after a heavy beating are over-bet, whatever the yard's record

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-53ab001f (Recruit to a yard good with new arrivals, after a heavy beating for it) was predicted positive but came out the other way: A/E 0.93 on 422 bets (p 0.399 for the opposite direction). O-0f113025
   • H-61d14772 (Negative: short-priced recruits to a strong new yard after a heavy bea) pointed the predicted way but failed the bar: A/E 0.81 on 84 bets (p 0.809), control 1.04. O-282acadf
   • H-432686d7 (Negative: short-priced planned moves from yards proven poor at that mo) pointed the predicted way but failed the bar: A/E 0.87 on 195 bets (p 0.82), control 1.05. O-bc2c08d8
2. AskedAre short-priced horses having their first run for a new trainer after a heavy beating over-bet whatever the new yard's record, i.e. does the market price the move itself rather than the yard? answered
3. MechanismA horse beaten heavily and then moved is backed by connections and punters who assume the new yard will fix it. The horse's ability has not changed, and a heavy beating often reflects a real limit. Money for the switch shortens the price beyond what the horse's form supports.
4. Why the market misses itThe trainer change is visible on every racecard and draws stable-switch money. The heavy beating is put down to the old yard instead of being read as a true measure of the horse, so the short price carries narrative more than form.
5. Predictednegative, A/E 0.82 to 0.92 on 300+ bets

Mechanism and rationale

A horse beaten heavily and then moved is backed by connections and punters who assume the new yard will fix it. The horse's ability has not changed, and a heavy beating often reflects a real limit. Money for the switch shortens the price beyond what the horse's form supports.

The rule

trainer_id <> last_trainer_id AND last_field_size >= 6 AND last_position >= 0.6 * last_field_size AND sp_decimal <= 8

Reads: trainer_id, last_trainer_id, last_field_size, last_position, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

ff4decec480d8581ace11424d1c77309b1cf20746da4a1ff5c916f24c2184189

Frozen

17 Oct, 05:13

Nothing in the definition can change after this.

Confounders

• Country mix: GB recruits after a beating scored 0.81 on 316 and IRE 1.35 on 106 in the strong-yard version. No country column exists, so report the split by course or region after the fact and do not filter on it
• Big-yard dispersal and owner-driven mass transfers, where many horses move at once for reasons that have nothing to do with the horse
• Field size and favourite status: short prices concentrate in small fields and among favourites, and the favourite-longshot bias works the other way. Check A/E within favourite and non-favourite subsets
• Seasonal clustering of transfers (end of season, sales periods) and switches between codes that come with a trainer change
• Price band inside the short range: the effect may sit entirely at SP 5.0 or shorter
• Short price after a bad run regardless of trainer, which the same-trainer comparison is there to separate out
• Last-run non-finishers are excluded, because last_position is NULL and the predicate drops them, so pulled-up runners are not in the set

Control

Discovery result

pass broad rule · 17 Oct, 05:11

Bets

400

65 winners

A/E

0.81

95%: 0.63 to 1.03

SP return

-31.2%

level stakes

Betfair SP return

-20.5%

380 bets with a BSP

Strike rate 16.2%. p = 0.96. Same-race control: the other runners in these races had A/E 1.05, so the selection did not 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 5 survived.

• control: Control: same trainer, same heavy beating, same short pricepass

  8,095 bets · A/E 1.00 (0.95 to 1.05) · control 1.00 · 17 Oct, 05:13

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-273,8611.04

  after 2026-04-274,2340.96

• perturbation: Perturb looser: beating cut 0.5, field 5+, SP 9.0 or shorterpass

  651 bets · A/E 0.94 (0.78 to 1.13) · control 1.01 · 17 Oct, 05:13

Sealed holdout (vault)

Bets

1,793

17 Oct, 05:13

A/E

1.02

p = 0.65

BH threshold then

0.0017

over every vault test so far

Outcome

Failed

Forward record

No forward record. Shadow tracking starts after the vault and counts only races run after the freeze.

Lineage

Agent runs

• designer v1ok · 52s · 17 Oct, 05:07

  Output (schema-checked)

  {
    "rule": "trainer_id <> last_trainer_id AND last_field_size >= 6 AND last_position >= 0.6 * last_field_size AND sp_decimal <= 8",
    "refinements": [
      {
        "label": "Shorter prices only (SP 5.0 or shorter): switch money should distort most where the price is shortest",
        "rule": "trainer_id <> last_trainer_id AND last_field_size >= 6 AND last_position >= 0.6 * last_field_size AND sp_decimal <= 5"
      },
      {
        "label": "Heavier beating (bottom quarter of a field of 8+): the more decisive the beating, the more it reflects a real limit that the new-yard story ignores",
        "rule": "trainer_id <> last_trainer_id AND last_field_size >= 8 AND last_position >= 0.75 * last_field_size AND sp_decimal <= 8"
      }
    ],
    "control": {
      "description": "Same-race control: the other runners in the same races. They should score close to 1.00 (earlier new-yard and planned-move rules had controls of 1.00 to 1.05). The effect only counts if the rule's A/E is at least 0.05 below the same-race control. A second comparison is reported for reference and is not part of the pass test: short-priced runners with the same heavy beating last time that stayed with the same trainer. This shows whether the over-betting comes from the trainer switch or simply from a short price after a bad run.",
      "rule": "trainer_id = last_trainer_id AND last_field_size >= 6 AND last_position >= 0.6 * last_field_size AND sp_decimal <= 8"
    },
    "pass_criteria": {
      "min_bets": 250,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 120,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Country mix: GB recruits after a beating scored 0.81 on 316 and IRE 1.35 on 106 in the strong-yard version. No country column exists, so report the split by course or region after the fact and do not filter on it",
      "Big-yard dispersal and owner-driven mass transfers, where many horses move at once for reasons that have nothing to do with the horse",
      "Field size and favourite status: short prices concentrate in small fields and among favourites, and the favourite-longshot bias works the other way. Check A/E within favourite and non-favourite subsets",
      "Seasonal clustering of transfers (end of season, sales periods) and switches between codes that come with a trainer change",
      "Price band inside the short range: the effect may sit entirely at SP 5.0 or shorter",
      "Short price after a bad run regardless of trainer, which the same-trainer comparison is there to separate out",
      "Last-run non-finishers are excluded, because last_position is NULL and the predicate drops them, so pulled-up runners are not in the set"
    ],
    "mechanism": "A horse beaten heavily and then moved to a new trainer is backed by connections and punters on the assumption that the new yard will fix it. The horse's ability has not changed, and a heavy beating (finishing in the bottom 40% of a field of 6 or more) often reflects a real limit. Money for the switch shortens the price beyond what the form supports. If the story is right, the over-betting comes from the visible move itself and not from the new yard's record: it showed up for strong-record yards (0.81 on 84) and for poor-record yards on other moves (0.87 on 195).",
    "market_miss": "The trainer change is printed on every racecard and draws stable-switch money. The heavy beating is put down to the old yard instead of being read as a true measure of the horse, so a short price on such a runner carries narrative more than form. Because this happens whatever the new yard's record, punters are pricing the switch itself and not any evidence about the yard.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.82 to 0.92 on 300+ bets, at least 0.05 below the same-race control",
      "falsified_if": "A/E at or above 0.97 on 250+ bets, or A/E not at least 0.05 below the same-race control; also counts against the switch mechanism if the same-trainer comparison is just as low"
    },
    "notes": "The rule sketch is kept as written. Every column is pre-off and available from mid-2020 in both countries. The rule generalises H-61d14772 by dropping the tx_new_yard record conditions, which answers Q-281e16e3 (is it the move or the yard?). Pass and forward criteria are read in the negative direction: the vault judges whether the runners are over-bet, i.e. the inverse ratio is at least 1.05. Min bets is raised to 250 to match the falsification clause. There is no country column, so the GB/IRE split cannot be a refinement and must be reported after the fact."
  }

Timeline

1. Job 3178 finished

   17 Oct, 05:13 · engine:compute-1

2. H-a584956c Vault → KILLEDfailed the vault vault CONTRADICTED: A/E 1.019 on 1793 bets, p 0.6508 against BH threshold 0.001724 over 58 vault tests

   17 Oct, 05:13 · engine:orchestrator

3. vault result for H-a584956c (vault query (contradicted)): fail

   17 Oct, 05:13 · engine:orchestrator

4. H-a584956c Frozen → FROZENvault vault query started

   17 Oct, 05:13 · engine:orchestrator

5. compute-1 took job 3178 (vault, attempt 1)

   17 Oct, 05:13 · compute-1

6. Job 3170 finished

   17 Oct, 05:13 · engine:compute-1

7. Queued vault on the compute lane: frozen: one sealed vault query

   17 Oct, 05:13 · engine:scheduler

8. H-a584956c Falsification → FROZEN survived falsification

   17 Oct, 05:13 · engine:falsification

9. falsification result for H-a584956c (partition: Partition: non-favourites only, to remove favourite-status confounding): pass

   17 Oct, 05:13 · engine:evaluator

10. compute-1 took job 3170 (attack, attempt 1)

    17 Oct, 05:13 · compute-1

11. Job 3169 finished

    17 Oct, 05:13 · engine:compute-1

12. falsification result for H-a584956c (partition: Partition: handicaps only, where a mark should already price in the beating): pass

    17 Oct, 05:13 · engine:evaluator

13. compute-1 took job 3169 (attack, attempt 1)

    17 Oct, 05:13 · compute-1

14. Job 3168 finished

    17 Oct, 05:13 · engine:compute-1

15. falsification result for H-a584956c (perturbation: Perturb tighter: beating cut 0.7, field 7+, SP 7.0 or shorter): pass

    17 Oct, 05:13 · engine:evaluator

16. compute-1 took job 3168 (attack, attempt 1)

    17 Oct, 05:13 · compute-1

17. Job 3167 finished

    17 Oct, 05:13 · engine:compute-1

18. falsification result for H-a584956c (perturbation: Perturb looser: beating cut 0.5, field 5+, SP 9.0 or shorter): pass

    17 Oct, 05:13 · engine:evaluator

19. compute-1 took job 3167 (attack, attempt 1)

    17 Oct, 05:13 · compute-1