Rival: the unseated edge is only longshot drift; capped at 14/1 it matches the price

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
   • H98-brought-down-or-unseated (H98: brought down or unseated) was GREEN in the earlier research: A/E 1.10 on 612 bets (periods 1.17 / 1.02); now PROMISING. O-9b549ef0
2. AskedIs the brought-down/unseated edge only a 16/1+ longshot effect, and is it fading (1.17 then 1.02)? answered
3. MechanismRunners coming off a non-finish drift to big prices. The A/E lift comes from the general underpricing of jump outsiders plus a handful of very big exchange winners, not from any misreading of the incident.
4. Why the market misses itIf the rival is right, nothing is mispriced at sensible prices. The +140% BSP ROI against +16.6% SP ROI is a few outliers, and any real edge sits where the price is fair.
5. Predictedpositive, Rival predicts A/E 0.97 to 1.03 on 350+ capped bets; the thread mechanism predicts 1.08 or more
6. Falsified ifThe capped rule scores A/E 1.08 or more with both periods above 1.03, which would rule the longshot rival out
7. Rule sketchlast_non_finish_code IN ('BD', 'UR') AND sp_decimal <= 15

Mechanism and rationale

Runners coming off a non-finish drift to big prices. The A/E lift comes from the general underpricing of jump outsiders plus a handful of very big exchange winners, not from any misreading of the incident.

The rule

last_non_finish_code IN ('BD', 'UR') AND sp_decimal <= 15

Reads: last_non_finish_code, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

e128da9757ade19f9fc2c0a4f430edea74a67ffd6e1b6b4c3a7fc8458923ce41

Frozen

not frozen

The definition can still change.

Confounders

• Period decline: the pooled edge fell from 1.17 to 1.02, so a capped result has to be read by period, not pooled
• Race-type mix: almost all BD/UR runners are jumpers, so the control must be jump runners at matched prices
• Field size: big fields push prices out and change how many runners fall under the 15.0 cap
• Coding window: last_non_finish_code starts 2025-01-01 and last_code starts 2025-11-01, so this sample is wider than the original 612-bet one and periods won't line up exactly
• The BSP-vs-SP gap: a few outsized exchange winners carried the uncapped profit, so ROI must be judged on SP inside the cap
• UR (rider error) and BD (another horse's fault) may behave differently; pooling them can hide a split

Control

{
  "description": "Same-race rivals should score about 1.0 at matched prices. The main control is jump runners in the same 14/1-or-shorter band whose last run ended in a finish. They share the price band and race mix but had no incident, which separates any edge from the general pricing of capped jump runners. If the BD/UR group beats the price but this control also scores about 1.05 or more, the lift is a property of the price band and race type, not of the incident.",
  "rule": "last_position IS NOT NULL AND prior_runs >= 1 AND sp_decimal <= 15 AND race_type IN ('Hurdle', 'Chase')"
}

Discovery result

fail broad rule · 16 Oct, 13:15

Bets

312

49 winners

A/E

1.02

95%: 0.76 to 1.35

SP return

-15.8%

level stakes

Betfair SP return

+3.3%

311 bets with a BSP

Strike rate 15.7%. p = 0.46. 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 · 38s · 16 Oct, 13:15

  Output (schema-checked)

  {
    "rule": "last_non_finish_code IN ('BD', 'UR') AND sp_decimal <= 15",
    "refinements": [
      {
        "label": "jumps only, capped",
        "rule": "last_non_finish_code IN ('BD', 'UR') AND sp_decimal <= 15 AND race_type IN ('Hurdle', 'Chase')"
      },
      {
        "label": "shorter-price band (7/1 or shorter)",
        "rule": "last_non_finish_code IN ('BD', 'UR') AND sp_decimal <= 8"
      },
      {
        "label": "capped, back within four weeks",
        "rule": "last_non_finish_code IN ('BD', 'UR') AND sp_decimal <= 15 AND days_since_last_run <= 28"
      }
    ],
    "control": {
      "description": "Same-race rivals should score about 1.0 at matched prices. The main control is jump runners in the same 14/1-or-shorter band whose last run ended in a finish. They share the price band and race mix but had no incident, which separates any edge from the general pricing of capped jump runners. If the BD/UR group beats the price but this control also scores about 1.05 or more, the lift is a property of the price band and race type, not of the incident.",
      "rule": "last_position IS NOT NULL AND prior_runs >= 1 AND sp_decimal <= 15 AND race_type IN ('Hurdle', 'Chase')"
    },
    "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": [
      "Period decline: the pooled edge fell from 1.17 to 1.02, so a capped result has to be read by period, not pooled",
      "Race-type mix: almost all BD/UR runners are jumpers, so the control must be jump runners at matched prices",
      "Field size: big fields push prices out and change how many runners fall under the 15.0 cap",
      "Coding window: last_non_finish_code starts 2025-01-01 and last_code starts 2025-11-01, so this sample is wider than the original 612-bet one and periods won't line up exactly",
      "The BSP-vs-SP gap: a few outsized exchange winners carried the uncapped profit, so ROI must be judged on SP inside the cap",
      "UR (rider error) and BD (another horse's fault) may behave differently; pooling them can hide a split"
    ],
    "mechanism": "Rival explanation: horses coming off an unseat or brought-down drift to big prices. Any A/E lift reflects the general underpricing of jump outsiders plus a handful of very big exchange winners, not a misreading of the incident. If so, capping at 14/1 (SP decimal 15.0 or shorter) should remove the lift. The thread's mechanism instead says the market wrongly treats a non-completion through no fault of the horse's ability as a form blot. Under that view the lift should survive at sensible prices.",
    "market_miss": "If the rival is right, nothing is mispriced at 14/1 or shorter, and the +140% BSP ROI against +16.6% SP ROI is just a few outliers. If the thread is right, the market marks down BD/UR runners even when they are fancied, because a letter in the form figures looks like failure.",
    "prediction": {
      "direction": "positive",
      "size": "Thread mechanism: A/E 1.08 or more on 350+ capped bets. Rival: 0.97 to 1.03.",
      "falsified_if": "For the thread: the capped rule scores below 1.03, or one period is at or below 1.0, which supports the longshot rival. For the rival: A/E 1.08 or more with both periods above 1.03."
    },
    "notes": "I used last_non_finish_code (from 2025-01-01) rather than last_code (from 2025-11-01) because it gives a longer window for the same BD/UR codes. Decimal 15.0 is 14/1. The vault reads in the positive direction, so passing means the rival is refuted. Report results by price band (up to 3/1, 4/1 to 7/1, 8/1 to 14/1) and by period, and give SP ROI next to A/E."
  }

Timeline

1. Job 1563 finished

   16 Oct, 13:15 · engine:compute-1

2. H-799d5af4 Discovery → KILLED fails: A/E 1.021, needs 1.05; p 0.4601, needs 0.05 or lower; periods [1.12, 0.9], each needs 1.0; A/E 1.021 against same-race control 0.999, needs 0.03 above (not a niche pass: A/E 1.021 below 1.05; does not beat the same-race control 0.999; a period points the other way [1.12, 0.9])

   16 Oct, 13:15 · engine:evaluator

3. discovery result for H-799d5af4 (broad rule): fail

   16 Oct, 13:15 · engine:evaluator

4. H-799d5af4 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 13:15 · engine:orchestrator

5. compute-1 took job 1563 (evaluate, attempt 1)

   16 Oct, 13:15 · compute-1

6. Job 1561 finished

   16 Oct, 13:15 · engine:ai-1

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

   16 Oct, 13:15 · engine:scheduler

8. H-799d5af4 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 13:15 · agent:designer@v1

9. ai-1 took job 1561 (design, attempt 1)

   16 Oct, 13:15 · ai-1

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

    16 Oct, 13:14 · engine:scheduler

11. New idea H-799d5af4: Rival: the unseated edge is only longshot drift; capped at 14/1 it matches the price (rival-explanation)

    16 Oct, 13:14 · 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.