Test-mechanism: quick-returning winners at 9/1+ whose last win came at 7/1+ are over-bet

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-9e92ea71 (Quick-returning last-time winners of any race running in a handicap at) pointed the predicted way but failed the bar: A/E 0.65 on 424 bets (p 0.97), control 1.02. O-3e924322
   • H-4f18f696 (Replicate: quick-returning last-time winners at 9/1+ in non-handicaps ) pointed the predicted way but failed the bar: A/E 0.46 on 106 bets (p 0.932), control 1.03. O-83cf6feb
2. AskedAmong quick-returning last-time winners at 9/1+, is the shortfall concentrated in horses whose previous win came at a long price, a surprise win the market now over-trusts? answered
3. MechanismA long-priced winner more often won through luck, a weak race, an uncontested lead or a pace collapse. That makes the win the kind the result flatters, and the next run regresses. When such a horse is turned out quickly, the market still treats the win as real form.
4. Why the market misses itThe form book shows a win, not the price the win came at or the circumstances behind it. Punters reprice the horse on the '1', while the original long price was the better estimate of its ability.
5. Predictednegative, A/E 0.55 to 0.80 on 250+ bets, below the 0.65 of the full quick-return pool
6. Falsified ifA/E 0.90 or higher on 200+ bets, or a deficit no deeper than among quick-returning winners whose last win came at under 7/1
7. Rule sketchlast_position = 1 AND sp_decimal > 9 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND last_sp_decimal IS NOT NULL AND last_sp_decimal > 7

The question's answer so far: No. Quick-returning winners whose last win came at 7/1+ scored 0.94 on 228, near par and much shallower than the 0.62 parent, so the shortfall is not concentrated in surprise winners.

Mechanism and rationale

A long-priced winner more often won through luck, a weak race, an uncontested lead or a pace collapse. That makes the win the kind the result flatters, and the next run regresses. When such a horse is turned out quickly, the market still treats the win as real form.

The rule

last_position = 1 AND sp_decimal > 9 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND last_sp_decimal IS NOT NULL AND last_sp_decimal > 7

Reads: last_position, sp_decimal, days_since_last_run, last_sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

dc0246c8bbdf5aa880097c5755a9dcf022fedfe58e1e9fd5d23f55b83a512272

Frozen

not frozen

The definition can still change.

Confounders

• Ordinary longshot bias: horses long-priced at both runs may just be poor, so the deficit could be the usual 16/1+ under-performance (the up-to-14/1 refinement checks this)
• Small-field or weak races that produced the surprise win (last_field_size, last_race_class)
• IRE versus GB market depth: earlier negative pools were thinner and deeper in IRE
• Code mix: Flat versus jumps quick returners differ in how often they turn out and how prices form
• Handicap mark rise after the win: earlier work found the mark is not the driver, but it is correlated with the win price
• Small sample: the full quick-return pool is 530, so the 7/1+ subset may come in under 250

Control

{
  "description": "Same-race other runners should sit near par (about 1.0, as in the earlier pools at 1.02), which shows the deficit is not a property of the races. The key mechanism comparison is quick-returning winners at 9/1+ whose last win came at 7/1 or shorter (an 'expected' win). If the mechanism holds, that group should be clearly less over-bet than the main rule. If it is as deep or deeper, the surprise-win explanation fails and the quick-return deficit is about something else.",
  "rule": "last_position = 1 AND sp_decimal > 9 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND last_sp_decimal IS NOT NULL AND last_sp_decimal <= 7"
}

Discovery result

fail broad rule · 17 Oct, 01:04

Bets

228

11 winners

A/E

0.94

95%: 0.47 to 1.68

SP return

-41.9%

level stakes

Betfair SP return

-26.8%

216 bets with a BSP

Strike rate 4.8%. p = 0.62. Same-race control: the other runners in these races had A/E 1.00, 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 · 35s · 17 Oct, 01:00

  Output (schema-checked)

  {
    "rule": "last_position = 1 AND sp_decimal > 9 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND last_sp_decimal IS NOT NULL AND last_sp_decimal > 7",
    "refinements": [
      {
        "label": "Up to 14/1 today: separates surprise-win regression from ordinary longshot bias",
        "rule": "last_position = 1 AND sp_decimal > 9 AND sp_decimal <= 15 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND last_sp_decimal IS NOT NULL AND last_sp_decimal > 7"
      },
      {
        "label": "Handicaps only: the setting where the quick-return winner deficit was established (0.651 on 424)",
        "rule": "last_position = 1 AND is_handicap = 1 AND sp_decimal > 9 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND last_sp_decimal IS NOT NULL AND last_sp_decimal > 7"
      }
    ],
    "control": {
      "description": "Same-race other runners should sit near par (about 1.0, as in the earlier pools at 1.02), which shows the deficit is not a property of the races. The key mechanism comparison is quick-returning winners at 9/1+ whose last win came at 7/1 or shorter (an 'expected' win). If the mechanism holds, that group should be clearly less over-bet than the main rule. If it is as deep or deeper, the surprise-win explanation fails and the quick-return deficit is about something else.",
      "rule": "last_position = 1 AND sp_decimal > 9 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND last_sp_decimal IS NOT NULL AND last_sp_decimal <= 7"
    },
    "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": [
      "Ordinary longshot bias: horses long-priced at both runs may just be poor, so the deficit could be the usual 16/1+ under-performance (the up-to-14/1 refinement checks this)",
      "Small-field or weak races that produced the surprise win (last_field_size, last_race_class)",
      "IRE versus GB market depth: earlier negative pools were thinner and deeper in IRE",
      "Code mix: Flat versus jumps quick returners differ in how often they turn out and how prices form",
      "Handicap mark rise after the win: earlier work found the mark is not the driver, but it is correlated with the win price",
      "Small sample: the full quick-return pool is 530, so the 7/1+ subset may come in under 250"
    ],
    "mechanism": "A win at 7/1 or longer is more often down to luck, a weak race, an uncontested lead or a pace collapse than a win at a short price. That makes it a win the result flatters, and the next run regresses toward the ability the original price implied. When the horse is turned out within 14 days, the win is the most recent and most visible form line, and the market prices it as real improvement.",
    "market_miss": "The form figure shows a '1', not the price the win came at or how it happened. Punters reprice the horse on the win, while the long price before that win was the better estimate of its ability. A quick return leaves no later run that could correct the impression.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.55 to 0.80 on 200+ bets, below the 0.65 of the full quick-return 9/1+ winner pool",
      "falsified_if": "A/E 0.90 or higher on 200+ bets, or the main rule's A/E no lower than the control group (quick-returning 9/1+ winners whose last win came at 7/1 or shorter)"
    },
    "notes": "I kept the rule sketch as written: every column is pre-off and none reads today's result. The sketch said 250+ bets, but I lowered that to 200+ in the prediction because the parent pool is only 530. Two things should be read together: the gap between the main rule and the control rule is the actual mechanism test, and the up-to-14/1 refinement guards against plain longshot bias. Non-handicaps alone are too thin to read (earlier finding), so I gave them no refinement of their own."
  }

Timeline

1. Job 2118 finished

   17 Oct, 01:04 · engine:compute-1

2. H-20b631ca Discovery → KILLED fails: 228 bets, needs 300; p 0.4968 (lower tail), needs 0.05; periods [0.61, 1.18], each needs 0.99 or lower

   17 Oct, 01:04 · engine:evaluator

3. discovery result for H-20b631ca (broad rule): fail

   17 Oct, 01:04 · engine:evaluator

4. H-20b631ca Designed → RESEARCHdiscovery measuring on the discovery zone

   17 Oct, 01:04 · engine:orchestrator

5. compute-1 took job 2118 (evaluate, attempt 1)

   17 Oct, 01:04 · compute-1

6. Job 2088 finished

   17 Oct, 01:00 · engine:ai-1

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

   17 Oct, 01:00 · engine:scheduler

8. H-20b631ca Proposed → RESEARCHdesigned pre-spec written and hashed

   17 Oct, 01:00 · agent:designer@v1

9. ai-1 took job 2088 (design, attempt 1)

   17 Oct, 01:00 · ai-1

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

    16 Oct, 17:03 · engine:scheduler

11. New idea H-20b631ca: Test-mechanism: quick-returning winners at 9/1+ whose last win came at 7/1+ are over-bet (test-mechanism)

    16 Oct, 17:03 · 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.