Rival: the chase unseater lift is a long-price artefact; test unseaters under 16/1

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-f8d6b9d7 (Generalise: all last-time chase unseaters beat their price, going well) passed discovery: A/E 1.4 on 213 bets (p 0.048), same-race control 0.96. O-4c8868a6
   • H-9f59a0ed (Sharpen: chase unseaters with no jumping error recorded last time) passed discovery as a niche: A/E 1.41 on 110 bets (p 0.131), same-race control 0.96. O-e6040866
2. AskedDo last-time chase unseaters beat their price below 16/1, or is the A/E 1.40 carried by 4 long-priced winners? answered
3. MechanismAn unseat often happens through bad luck, interference or a rider losing balance, and it tells little about the horse's ability. If the market over-punishes the UR letter, horses at every price should beat their odds. If the lift is only the noise of a few 16/1+ winners, it vanishes below 16/1.
4. Why the market misses itPunters read UR as a failed run and drift the horse. The figures say the horse was usually competing normally when it lost its rider.
5. Predictedpositive, A/E 1.10 to 1.25 on 110+ bets
6. Falsified ifA/E at or below 1.02 under 16/1, which would mean the unseat edge is long-shot variance and not a mispriced letter.
7. Rule sketch

Mechanism and rationale

An unseat often happens through bad luck, interference or a rider losing balance, and it tells little about the horse's ability. If the market over-punishes the UR letter, horses at every price should beat their odds. If the lift is only the noise of a few 16/1+ winners, it vanishes below 16/1.

The rule

race_type = 'Chase' AND last_non_finish_code = 'UR' AND sp_decimal < 17

Reads: race_type, last_non_finish_code, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

f0efa00a361ff8d39658332bdb268d02fb5900b4656244cae5b97017fffa7231

Frozen

17 Oct, 03:09

Nothing in the definition can change after this.

Confounders

• Field size and race type at long prices: excluding 16/1+ removes most big-field handicap longshots, which changes the race mix as well as the price
• Small number of winners: with roughly 120 bets under 16/1 and about 20 winners, one or two results move A/E by 0.1
• Handicap versus non-handicap mix (refinement 2 addresses this)
• Same-race favourite-longshot bias: the price cut alone pulls A/E up relative to an unrestricted sample, so judge it against the same-race and PU controls at matching prices
• Column window: last_non_finish_code only separates UR from PU from 2025-01-01 (every earlier non-finisher is coded PU), so the sample is about two seasons, and a vault window before 2025 will return 0 bets, as happened with H-f8d6b9d7
• Overlap with the going-well system: some unseaters are flagged going-well, and that lift has its own price profile

Control

{
  "description": "The lab's default: other runners in the same races. Expect A/E about 0.95 to 1.00, in line with the 0.96 to 0.97 controls seen on the broad unseater cuts. The second control is chase runners under 16/1 whose last run ended in a pull-up, a non-finish that does say something about ability. It uses only the window where codes are real (from 2025). If unseaters beat this group at the same prices, the edge belongs to the UR letter and not to non-finishers in general. Expect PU runners near or below 1.00.",
  "rule": "race_type = 'Chase' AND last_non_finish_code = 'PU' AND sp_decimal < 17 AND race_date >= '2025-01-01'"
}

Discovery result

pass broad rule (niche) · 17 Oct, 03:02

Bets

122

25 winners

A/E

1.35

95%: 0.88 to 2.00

SP return

+19.9%

level stakes

Betfair SP return

+40.8%

122 bets with a BSP

Strike rate 20.5%. p = 0.084. Same-race control: the other runners in these races had A/E 0.94, so the selection 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 6 survived.

• price_band: Short-price core: 7/1 or shorter onlypass

  62 bets · A/E 1.20 (0.70 to 1.93) · control 0.94 · 17 Oct, 03:09

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27461.33

  after 2026-04-2716–

• perturbation: Perturb price cut down 25%: under 12/1pass

  101 bets · A/E 1.33 (0.84 to 1.99) · control 0.93 · 17 Oct, 03:09

  Breakdown

Sealed holdout (vault)

Bets

0

17 Oct, 03:09

A/E

–

p = 1.0

BH threshold then

0.0020

over every vault test so far

Outcome

Failed

Forward record

• forward vs bar {'min_bets': 100, 'min_ae': 1.05, 'max_p': 0.2}info

  0 bets · · 17 Oct, 03:10

  Breakdown

  betsA/E0.71.01.3

  forward0–

Other experiments

• neighbour: < 17 -> 14 (-20%)pass

  114 bets · A/E 1.33 (0.85 to 1.98) · control 0.94 · 17 Oct, 03:02

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27721.35

  after 2026-04-27421.27

• neighbour: < 17 -> 20 (+20%)pass

  146 bets · A/E 1.34 (0.88 to 1.96) · control 0.95 · 17 Oct, 03:02

  Breakdown

Lineage

Agent runs

• designer v1ok · 40s · 17 Oct, 03:00

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND last_non_finish_code = 'UR' AND sp_decimal < 17",
    "refinements": [
      {
        "label": "8/1 or shorter: the lift should survive at prices where long-shot variance cannot carry it",
        "rule": "race_type = 'Chase' AND last_non_finish_code = 'UR' AND sp_decimal < 17 AND sp_decimal <= 9"
      },
      {
        "label": "Handicaps only: removes the handicap/non-handicap mix confounder, where most unseaters run",
        "rule": "race_type = 'Chase' AND last_non_finish_code = 'UR' AND sp_decimal < 17 AND is_handicap = 1"
      }
    ],
    "control": {
      "description": "The lab's default: other runners in the same races. Expect A/E about 0.95 to 1.00, in line with the 0.96 to 0.97 controls seen on the broad unseater cuts. The second control is chase runners under 16/1 whose last run ended in a pull-up, a non-finish that does say something about ability. It uses only the window where codes are real (from 2025). If unseaters beat this group at the same prices, the edge belongs to the UR letter and not to non-finishers in general. Expect PU runners near or below 1.00.",
      "rule": "race_type = 'Chase' AND last_non_finish_code = 'PU' AND sp_decimal < 17 AND race_date >= '2025-01-01'"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1.05,
      "max_p": 0.2
    },
    "confounders": [
      "Field size and race type at long prices: excluding 16/1+ removes most big-field handicap longshots, which changes the race mix as well as the price",
      "Small number of winners: with roughly 120 bets under 16/1 and about 20 winners, one or two results move A/E by 0.1",
      "Handicap versus non-handicap mix (refinement 2 addresses this)",
      "Same-race favourite-longshot bias: the price cut alone pulls A/E up relative to an unrestricted sample, so judge it against the same-race and PU controls at matching prices",
      "Column window: last_non_finish_code only separates UR from PU from 2025-01-01 (every earlier non-finisher is coded PU), so the sample is about two seasons, and a vault window before 2025 will return 0 bets, as happened with H-f8d6b9d7",
      "Overlap with the going-well system: some unseaters are flagged going-well, and that lift has its own price profile"
    ],
    "mechanism": "An unseat usually comes from bad luck, interference, a peck, or a rider losing balance, and often happens while the horse is competing normally. So the UR letter says little about the horse's ability or form. If the market drifts every last-time unseater on the strength of the letter alone, the mispricing should show at all prices, including 16/1 and shorter. If the 1.40 A/E on all chase unseaters is only the noise of 4 winners at 16/1+ (A/E 1.79 on 91), the edge should vanish below 16/1.",
    "market_miss": "Punters and form summaries read UR like any other failed run and drift the horse next time. They do not separate a lost rider, which usually happened during a normal run, from a pull-up or a tired fall, which do reflect the horse. If the drift also shows in the mid and short price range, where the market is most efficient, that would indicate a true letter-based mispricing.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.10 to 1.25 on roughly 110 to 130 bets at 16/1 or shorter (sp_decimal < 17)",
      "falsified_if": "A/E at or below 1.02 under 16/1, or no better than the matched-price PU control. Either would mean the chase unseater lift is long-shot variance carried by the 16/1+ winners, not a mispriced UR letter."
    },
    "notes": "sp_decimal < 17 means 16/1 or shorter, the exact complement of the '16/1 or bigger' cut (91 bets, A/E 1.79) in the findings, so the two halves add up to the 213-bet parent. The expected sample of about 122 bets is below the 150-bet discovery floor. A result that points the right way is expected to go on to falsification and forward tracking, not to pass discovery outright. The vault and forward windows must start on or after 2025-01-01, because last_non_finish_code cannot tell UR from PU before then."
  }

Timeline

1. Job 2599 finished

   17 Oct, 03:10 · engine:compute-1

2. forward result for H-a44d6bdc (forward vs bar {'min_bets': 100, 'min_ae': 1.05, 'max_p': 0.2}): info

   17 Oct, 03:10 · engine:evaluator

3. compute-1 took job 2599 (forward_check, attempt 1)

   17 Oct, 03:10 · compute-1

4. Queued forward_check on the compute lane: daily forward check of a shadow system

   17 Oct, 03:10 · engine:scheduler

5. Job 2582 finished

   17 Oct, 03:09 · engine:compute-1

6. H-a44d6bdc Vault → SHADOW vault SUPPORTIVE: A/E None on 0 bets, p 1 against BH threshold 0.002041 over 49 vault tests

   17 Oct, 03:09 · engine:orchestrator

7. vault result for H-a44d6bdc (vault query (supportive)): pass

   17 Oct, 03:09 · engine:orchestrator

8. H-a44d6bdc Frozen → FROZENvault vault query started

   17 Oct, 03:09 · engine:orchestrator

9. compute-1 took job 2582 (vault, attempt 1)

   17 Oct, 03:09 · compute-1

10. Job 2581 finished

    17 Oct, 03:09 · engine:compute-1

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

    17 Oct, 03:09 · engine:scheduler

12. H-a44d6bdc Falsification → FROZEN survived falsification

    17 Oct, 03:09 · engine:falsification

13. falsification result for H-a44d6bdc (partition: Handicaps only): pass

    17 Oct, 03:09 · engine:evaluator

14. compute-1 took job 2581 (attack, attempt 1)

    17 Oct, 03:09 · compute-1

15. Job 2580 finished

    17 Oct, 03:09 · engine:compute-1

16. falsification result for H-a44d6bdc (partition: Strip the going-well overlap): pass

    17 Oct, 03:09 · engine:evaluator

17. compute-1 took job 2580 (attack, attempt 1)

    17 Oct, 03:09 · compute-1

18. Job 2579 finished

    17 Oct, 03:09 · engine:compute-1

19. falsification result for H-a44d6bdc (control: Control: chase fallers proper at the same prices): pass

    17 Oct, 03:09 · engine:evaluator