Rival: chase-error lift is an 8/1 to 14/1 price-band effect, not a drift effect

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-e6eaa9ed (Rival: one-off chase-error edge is carried by horses fancied last time): the country 'GB' carries the effect (A/E 1.29) while the rest score IRE 0.98. O-59060787
   • H-e6eaa9ed (Rival: one-off chase-error edge is carried by horses fancied last time) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query. O-9bc242b7
   • H-e6eaa9ed (Rival: one-off chase-error edge is carried by horses fancied last time) passed discovery as a niche: A/E 1.22 on 204 bets (p 0.135), same-race control 0.97. O-d1ea467b
2. AskedDoes the one-off chase-error lift sit in the 8/1 to 14/1 band today, whatever the drift, so that drift only helps by moving runners into that band? answered
3. MechanismAfter a single blunder and a heavy defeat, a capable chaser is pushed out to a mid-market price. The market prices it off the bad figure and still treats it as a contender. Runners still at 3/1 or shorter are well backed by informed money and correctly priced. Runners at 16/1 or bigger are mostly genuinely outclassed. The mispricing would sit in the middle, whether the horse reached that price by drifting or not.
4. Why the market misses itAt 8/1 to 14/1 the price is set mainly by the form figures and the beaten distance in the last run. Stable or informed money is thin there. The in-running comment showing a one-off error is not reflected in mid-market prices.
5. Predictedpositive, A/E 1.40 to 1.65 on about 170 bets, with drifted and shortened runners inside the band within 0.25 of each other

Mechanism and rationale

After a single blunder and a heavy defeat, a capable chaser is pushed out to a mid-market price. The market prices it off the bad figure and still treats it as a contender. Runners still at 3/1 or shorter are well backed by informed money and correctly priced. Runners at 16/1 or bigger are mostly genuinely outclassed. The mispricing would sit in the middle, whether the horse reached that price by drifting or not.

The rule

race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10) AND sp_decimal >= 9 AND sp_decimal <= 15

Reads: race_type, remarked_last3, jumping_errors_last3, last_jumping_error, last_position, last_beaten_lengths, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

27619ba40d373efad40d60284c6e3d73a69cbfa3b39f456c812881460533d11b

Frozen

17 Oct, 01:53

Nothing in the definition can change after this.

Confounders

• Price-band slicing after seeing band results: the 8/1 to 14/1 band was chosen from earlier partitions, so the discovery A/E is optimistic. Out-of-sample forward tracking is the real test.
• Overlap: shares most runners with the drift (H-373590f6) and under-16/1 shadow rules, so it is not independent confirmation.
• General bounce-back: well-beaten fancied chasers with no jumping error scored 1.11, so part of any lift is not specific to the error.
• GB/IRE mix: earlier cuts were GB-led (IRE near par), and the IRE share may differ by band.
• One season only: remarked_last3 is available from November 2025, and late-season period skew has been seen in sibling cuts.
• Handicap share is higher at mid prices, so the band may be standing in for handicap status.
• Drift composition: drifted runners may be over-represented in the band, which would confound price level with drift unless the drift refinements are compared.

Control

{
  "description": "Two comparisons. (1) The other runners in the same races should sit near par (about 0.95 to 1.05), so the lift is not a property of mid-market chase races. (2) The same one-off-error, well-beaten chase pool outside the band (shorter than 8/1 or bigger than 14/1) should come in clearly below the in-band A/E, near par at the short end and at or below par at 16/1+. If the out-of-band pool matches the in-band A/E, price level is not what locates the edge. Inside the band, the drifted and shortened/steady refinements must be within 0.25 of each other. Drifted beating shortened by more than 0.3 falsifies the claim.",
  "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10) AND (sp_decimal < 9 OR sp_decimal > 15)"
}

Discovery result

pass broad rule (niche) · 17 Oct, 01:46

Bets

159

16 winners

A/E

1.46

95%: 0.83 to 2.37

SP return

+18.9%

level stakes

Betfair SP return

+50.4%

158 bets with a BSP

Strike rate 10.1%. p = 0.090. Same-race control: the other runners in these races had A/E 0.97, 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. 6 of 6 survived.

• perturbation: Perturb price band: shift to 7/1-12/1 (sp 7.5 to 13)pass

  180 bets · A/E 1.56 (0.99 to 2.35) · control 0.95 · 17 Oct, 01:53

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27771.21

  after 2026-04-271031.86

• perturbation: Perturb price band: widen to 7/1-16/1 (sp 8 to 17)pass

  213 bets · A/E 1.44 (0.90 to 2.18) · control 0.97 · 17 Oct, 01:53

Sealed holdout (vault)

Not queried. Only a frozen idea can spend its one query against the 2020-25 holdout.

Forward record

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

  0 bets · · 17 Oct, 01:53

  Breakdown

  betsA/E0.71.01.3

  forward0–

Other experiments

• neighbour: > 10 -> 8 (-20%)pass

  169 bets · A/E 1.37 (0.78 to 2.23) · control 0.98 · 17 Oct, 01:46

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-27651.35

  after 2026-04-271041.38

• neighbour: > 10 -> 12 (+20%)pass

  150 bets · A/E 1.56 (0.89 to 2.53) · control 0.97 · 17 Oct, 01:46

  Breakdown

Lineage

Agent runs

• designer v1ok · 33s · 17 Oct, 01:46

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10) AND sp_decimal >= 9 AND sp_decimal <= 15",
    "refinements": [
      {
        "label": "In-band, drifted since last run",
        "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10) AND sp_decimal >= 9 AND sp_decimal <= 15 AND last_sp_decimal IS NOT NULL AND sp_decimal > last_sp_decimal"
      },
      {
        "label": "In-band, same price or shortened since last run",
        "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10) AND sp_decimal >= 9 AND sp_decimal <= 15 AND last_sp_decimal IS NOT NULL AND sp_decimal <= last_sp_decimal"
      },
      {
        "label": "In-band, GB only",
        "rule": "region = 'GB' AND race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10) AND sp_decimal >= 9 AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "Two comparisons. (1) The other runners in the same races should sit near par (about 0.95 to 1.05), so the lift is not a property of mid-market chase races. (2) The same one-off-error, well-beaten chase pool outside the band (shorter than 8/1 or bigger than 14/1) should come in clearly below the in-band A/E, near par at the short end and at or below par at 16/1+. If the out-of-band pool matches the in-band A/E, price level is not what locates the edge. Inside the band, the drifted and shortened/steady refinements must be within 0.25 of each other. Drifted beating shortened by more than 0.3 falsifies the claim.",
      "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10) AND (sp_decimal < 9 OR sp_decimal > 15)"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.2,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1.05,
      "max_p": 0.2
    },
    "confounders": [
      "Price-band slicing after seeing band results: the 8/1 to 14/1 band was chosen from earlier partitions, so the discovery A/E is optimistic. Out-of-sample forward tracking is the real test.",
      "Overlap: shares most runners with the drift (H-373590f6) and under-16/1 shadow rules, so it is not independent confirmation.",
      "General bounce-back: well-beaten fancied chasers with no jumping error scored 1.11, so part of any lift is not specific to the error.",
      "GB/IRE mix: earlier cuts were GB-led (IRE near par), and the IRE share may differ by band.",
      "One season only: remarked_last3 is available from November 2025, and late-season period skew has been seen in sibling cuts.",
      "Handicap share is higher at mid prices, so the band may be standing in for handicap status.",
      "Drift composition: drifted runners may be over-represented in the band, which would confound price level with drift unless the drift refinements are compared."
    ],
    "mechanism": "After a single jumping error and a heavy defeat (beaten more than 10 lengths or failed to finish), a capable chaser whose other two recent runs had no error is priced off the bad figure. At 8/1 to 14/1 the market still sees it as an outside contender but underrates its chance of returning to form. At 3/1 or shorter, informed money prices it correctly. At 16/1 or bigger, most runners are genuinely outclassed. The claim is that the edge depends on today's price level, not on whether the horse drifted to get there.",
    "market_miss": "Mid-market prices are set mainly by the form figures and the last beaten distance, and stable or informed money is thin there. The in-running comment showing a one-off error (with a clean record otherwise) is not reflected in 8/1 to 14/1 prices, whereas short-priced runners attract enough informed money to correct it.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.40 to 1.65 on roughly 150 to 170 bets, with drifted and shortened/steady in-band runners within 0.25 of each other",
      "falsified_if": "In-band A/E of 1.20 or lower, or in-band drifted runners beating shortened/steady ones by more than 0.3 A/E (drift, not price level, carries the edge), or the out-of-band control matching the in-band A/E."
    },
    "notes": "I changed the rule sketch's sp_decimal > 8 to sp_decimal >= 9 so the band matches the stated 8/1 to 14/1. The sketch's version let in 15/2 (8.5), which is outside the stated band. 14/1 (15.0) stays included. I raised the discovery min_ae to 1.20 to match the hypothesis's own falsifier. The drift refinements need last_sp_decimal to be present. The sample is thin and taken from one season, so a positive result here is a lead and needs forward confirmation."
  }

Timeline

1. Job 2371 finished

   17 Oct, 01:53 · engine:compute-1

2. forward result for H-c66cb8b5 (forward vs bar {'min_bets': 100, 'min_ae': 1.05, 'max_p': 0.2}): info

   17 Oct, 01:53 · engine:evaluator

3. compute-1 took job 2371 (forward_check, attempt 1)

   17 Oct, 01:53 · compute-1

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

   17 Oct, 01:53 · engine:scheduler

5. Job 2364 finished

   17 Oct, 01:53 · engine:compute-1

6. H-c66cb8b5 Frozen → SHADOW vault skipped: remarked_last3, jumping_errors_last3, last_jumping_error, last_beaten_lengths do not exist before the discovery zone; track it forward instead

   17 Oct, 01:53 · engine:orchestrator

7. compute-1 took job 2364 (vault, attempt 1)

   17 Oct, 01:53 · compute-1

8. Job 2363 finished

   17 Oct, 01:53 · engine:compute-1

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

   17 Oct, 01:53 · engine:scheduler

10. H-c66cb8b5 Falsification → FROZEN survived falsification

    17 Oct, 01:53 · engine:falsification

11. falsification result for H-c66cb8b5 (partition: Partition: finished last time (heavy defeat, not a non-completion)): pass

    17 Oct, 01:53 · engine:evaluator

12. compute-1 took job 2363 (attack, attempt 1)

    17 Oct, 01:53 · compute-1

13. Job 2362 finished

    17 Oct, 01:53 · engine:compute-1

14. falsification result for H-c66cb8b5 (partition: Partition: handicap chases only): pass

    17 Oct, 01:53 · engine:evaluator

15. compute-1 took job 2362 (attack, attempt 1)

    17 Oct, 01:53 · compute-1

16. Job 2361 finished

    17 Oct, 01:53 · engine:compute-1

17. falsification result for H-c66cb8b5 (control: Control: same band, well-beaten chasers with no recent jumping error): pass

    17 Oct, 01:53 · engine:evaluator

18. compute-1 took job 2361 (attack, attempt 1)

    17 Oct, 01:53 · compute-1

19. Job 2360 finished

    17 Oct, 01:53 · engine:compute-1