Rival: the 8/1 to 14/1 chase lift is generic to well-beaten chasers with no jumping error

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-b63a044f (Rival: drift edge is a generic chase drifter effect after any well-bea) pointed the predicted way but failed the bar: A/E 1.09 on 658 bets (p 0.31), control 1.0. O-77afe601
   • H-c66cb8b5 (Rival: chase-error lift is an 8/1 to 14/1 price-band effect, not a dri) passed discovery as a niche: A/E 1.46 on 159 bets (p 0.09), same-race control 0.97. O-9d680f79
2. AskedDo well-beaten chasers with no jumping error last time, priced 8/1 to 14/1 today, score as well as the one-off-error pool in the same band (1.46)? answered
3. MechanismRival cause: any chaser beaten more than 10 lengths last time that the market still prices as a mid-range contender (8/1 to 14/1) bounces back more often than its price implies. The bad last figure is overweighted whatever caused it. On this account the jumping error plays no part.
4. Why the market misses itThe form figure and beaten distance drive mid-market prices for every well-beaten chaser. If that is the miss, it applies without a jumping-error remark, and the error story only marks a subset of it.
5. Predictedpositive, If the rival holds: A/E 1.30 or higher on 800+ bets. If the error story holds: A/E 1.00 to 1.12.
6. Falsified ifThe rival is ruled out if the error-free band pool scores 1.12 or lower while the error pool holds 1.46. The rival is ruled in if it scores 1.30 or higher with both periods above 1.1.

Mechanism and rationale

Rival cause: any chaser beaten more than 10 lengths last time that the market still prices as a mid-range contender (8/1 to 14/1) bounces back more often than its price implies. The bad last figure is overweighted whatever caused it. On this account the jumping error plays no part.

The rule

race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 0 AND last_position IS NOT NULL AND last_beaten_lengths > 10 AND sp_decimal >= 9 AND sp_decimal <= 15

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

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

f9345224e08644138e2c9f92a30ae64f0a43f0dc967380f86fd50e0ae6f04090

Frozen

not frozen

The definition can still change.

Confounders

• Band (9 to 15 decimal) was chosen after seeing earlier partitions of the error pool, and the edges are arbitrary, so in-band figures are optimistic for both pools
• This rule drops last-time non-finishers, which the error pool keeps (last_position IS NULL), so the mix differs: no pulled-up or brought-down horses here
• Handicap share is higher at mid prices, and handicap marks falling after a heavy defeat could drive a bounce-back regardless of cause
• Late-season skew (after 27 April 2026) seen in sibling cuts: check both periods
• GB/IRE mix differs from the error pool, and the thread's lifts have been GB-led
• General well-beaten-chaser bounce-back (1.11 for fancied error-free chasers) may explain part of the lift without being specific to the 8/1 to 14/1 band

Control

{
  "description": "Same-race other runners should sit near par (about 0.97, as in the error-pool band tests), so any lift is not a property of these races. The direct comparison the question asks for is the one-off-error pool in the same 9 to 15 decimal band, measured over the same window. The rival needs the error-free pool to come within about 0.15 of it (1.30 or higher against about 1.46). The error story needs the error-free pool to stay at 1.12 or lower while the error pool holds.",
  "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"
}

Discovery result

fail broad rule · 17 Oct, 01:53

Bets

302

24 winners

A/E

1.18

95%: 0.76 to 1.75

SP return

-11.8%

level stakes

Betfair SP return

+12.4%

300 bets with a BSP

Strike rate 7.9%. p = 0.24. Same-race control: the other runners in these races had A/E 0.99, 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 · 43s · 17 Oct, 01:51

  Output (schema-checked)

  {
    "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 0 AND last_position IS NOT NULL AND last_beaten_lengths > 10 AND sp_decimal >= 9 AND sp_decimal <= 15",
    "refinements": [
      {
        "label": "GB only",
        "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 0 AND last_position IS NOT NULL AND last_beaten_lengths > 10 AND sp_decimal >= 9 AND sp_decimal <= 15 AND region = 'GB'"
      },
      {
        "label": "Handicaps only (tests the handicap-share confounder)",
        "rule": "race_type = 'Chase' AND remarked_last3 = 3 AND jumping_errors_last3 = 0 AND last_position IS NOT NULL AND last_beaten_lengths > 10 AND sp_decimal >= 9 AND sp_decimal <= 15 AND is_handicap = 1"
      }
    ],
    "control": {
      "description": "Same-race other runners should sit near par (about 0.97, as in the error-pool band tests), so any lift is not a property of these races. The direct comparison the question asks for is the one-off-error pool in the same 9 to 15 decimal band, measured over the same window. The rival needs the error-free pool to come within about 0.15 of it (1.30 or higher against about 1.46). The error story needs the error-free pool to stay at 1.12 or lower while the error pool holds.",
      "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"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.12,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 150,
      "min_ae": 1.05,
      "max_p": 0.2
    },
    "prediction": {
      "direction": "positive",
      "size": "If the rival holds: A/E 1.30 or higher on 800+ bets, with both periods above 1.10. If the error story holds: A/E 1.00 to 1.12.",
      "falsified_if": "The rival is ruled out if the error-free band pool scores 1.12 or lower while the same-band error pool stays near 1.46. It is also not supported if it beats 1.12 only in the period after 27 April 2026, or if it lands between 1.12 and 1.30 (a partial generic bounce-back, as the 1.11 for fancied error-free chasers suggested)."
    },
    "confounders": [
      "Band (9 to 15 decimal) was chosen after seeing earlier partitions of the error pool, and the edges are arbitrary, so in-band figures are optimistic for both pools",
      "This rule drops last-time non-finishers, which the error pool keeps (last_position IS NULL), so the mix differs: no pulled-up or brought-down horses here",
      "Handicap share is higher at mid prices, and handicap marks falling after a heavy defeat could drive a bounce-back regardless of cause",
      "Late-season skew (after 27 April 2026) seen in sibling cuts: check both periods",
      "GB/IRE mix differs from the error pool, and the thread's lifts have been GB-led",
      "General well-beaten-chaser bounce-back (1.11 for fancied error-free chasers) may explain part of the lift without being specific to the 8/1 to 14/1 band"
    ],
    "mechanism": "Rival cause: a chaser beaten more than 10 lengths last time, but still priced as a mid-range contender today (decimal 9 to 15), bounces back more often than its price implies. The market overweights the bad last figure and beaten distance whatever caused them. On this account the jumping error is incidental: error-free well-beaten chasers in the same band should show a lift close to the one-off-error pool's 1.46.",
    "market_miss": "Mid-market prices for well-beaten chasers lean on the form figure and beaten distance. If that is the miss, it shows without any jumping-error remark, and the error story only flags a subset of a wider mispricing. If error-free runners sit at par, the miss is specific to how the market reads a one-off error.",
    "notes": "I kept the rule sketch unchanged. It uses only pre-off columns, as the sibling error-pool rules do, and it does not read the result. The discovery bar is stricter than the lab floor (300 bets, A/E 1.12), because the prediction expects 800+ bets and 1.12 is the hypothesis's own line for ruling the rival out. Refinements are GB only (the thread's results have been GB-led) and handicaps only (the handicap-share confounder). They are read only if the broad rule shows evidence. Read the result against the same-band error-pool control, not against par alone."
  }

Timeline

1. Job 2352 finished

   17 Oct, 01:53 · engine:compute-1

2. H-3cc36f8a Discovery → KILLED fails: p 0.2367, needs 0.05 or lower; periods [0.92, 1.42], each needs 1.0 (not a niche pass: a period points the other way [0.92, 1.42])

   17 Oct, 01:53 · engine:evaluator

3. discovery result for H-3cc36f8a (broad rule): fail

   17 Oct, 01:53 · engine:evaluator

4. H-3cc36f8a Designed → RESEARCHdiscovery measuring on the discovery zone

   17 Oct, 01:53 · engine:orchestrator

5. compute-1 took job 2352 (evaluate, attempt 1)

   17 Oct, 01:53 · compute-1

6. Job 2351 finished

   17 Oct, 01:52 · engine:ai-1

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

   17 Oct, 01:52 · engine:scheduler

8. H-3cc36f8a Proposed → RESEARCHdesigned pre-spec written and hashed

   17 Oct, 01:52 · agent:designer@v1

9. ai-1 took job 2351 (design, attempt 1)

   17 Oct, 01:51 · ai-1

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

    17 Oct, 01:51 · engine:scheduler

11. New idea H-3cc36f8a: Rival: the 8/1 to 14/1 chase lift is generic to well-beaten chasers with no jumping error (rival-explanation)

    17 Oct, 01:51 · 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.