S3: chase not completed large negative

Mechanism and rationale

Shock test S3-chase-not-completed-large-negative (shock.py, frozen before the holdout was fetched)

The rule

code = 'Chase' AND finished = 0 AND s <= -1.5

Reads: code, finished, s

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

2bde6ddc55e333b65ea9a11201ec31aaab1ddc49184e9699039bab726f24d4ec

Frozen

14 Oct, 01:00

Nothing in the definition can change after this.

Direction

positive

Frozen in

research/experiments/raceiq/shock_tests

Frozen on

2026-10-14

Id

S3-chase-not-completed-large-negative

Rule

code = 'Chase' AND finished = 0 AND s <= -1.5

Result (from the archive)

info discovery year (reference) · 15 Oct, 23:10

Bets

214

A/E

1.01

winners vs the prices

SP return

-12.1%

level stakes

Betfair SP return

+33.5%

213 bets with a BSP

Strike rate 8.4%.

Falsification

No attacks yet. Only promising ideas are attacked.

Sealed holdout (vault)

Bets

254

14 Oct, 13:00

A/E

1.09

p = 0.36

BH threshold then

0.0034

over every vault test so far

Outcome

Failed

from the archive

Forward record

• forward check refusederror

  · 15 Oct, 23:21

Lineage

Agent runs

• investigator v1ok · 34s · 17 Oct, 02:41

  Output (schema-checked)

  {
    "reflection": "This thread has one test so far. The rule is: the horse's last run was a chase, it did not finish, and its Jump Index came out at least 1.5 s.d. below expectation (code='Chase' AND finished=0 AND s<=-1.5). In discovery the rule was flat: A/E 1.007 on 214 bets. On the sealed holdout it pointed the right way, 1.095 on 254 bets, but p was 0.364 against a false-discovery threshold of 0.0034. That is noise-level, so the rule sits in SHADOW. The holdout lift is GB-only (GB 1.117, IRE 0.971). That matches the lab's chase-error findings, where non-finisher edges are GB-only and thin, and runners at 16/1 or bigger are at or below par in every cut. So we have not learned that the shock measure adds anything. The open issue is whether s carries information beyond the plain fact that the horse failed to complete a chase. The market sees that 'not completed' letter; it may not see how far below expectation the run was before it ended. If s adds nothing, the rule is just one more 'chase non-finisher' cut, and those have not worked.",
    "answers": [],
    "questions": [
      {
        "question": "Among last-time chase non-finishers, does a deeper jumping shock (more negative s) lead to a higher A/E next time, or is s irrelevant once the horse failed to finish?",
        "rationale": "The holdout lift (1.095 on 254) could belong to all chase non-finishers rather than to the shock. A dose-response check would read s, finished, code and sp_decimal, comparing A/E across s bands (<=-2.5, -2.5 to -1.5, -1.5 to -0.5, >-0.5). If there is no gradient, the shock family adds nothing for non-finishers and this thread can be closed.",
        "observation_ids": [
          "O-ddd7ee0a"
        ],
        "priority": 0.55
      },
      {
        "question": "Is the GB/IRE split (1.117 vs 0.971) in shocked chase non-finishers a real market difference or small-sample noise, given other chase-error cuts show the same split?",
        "rationale": "The same GB-only pattern appears in the one-off-error non-finisher finding (GB 1.588, IRE 0.433). A test would read region, code, finished, s, last_non_finish_code and sp_decimal, and would need forward bets from both countries over a second winter.",
        "observation_ids": [
          "O-ddd7ee0a",
          "O-69d892e9"
        ],
        "priority": 0.35
      }
    ],
    "hypotheses": [
      {
        "title": "Rival: chase non-finishers with only a mild shock score the same as large-shock ones",
        "family": "shock",
        "operator": "rival-explanation",
        "question_id": null,
        "new_question": 0,
        "parent_id": "S3-chase-not-completed-large-negative",
        "reasoning": {
          "observation_ids": [
            "O-ddd7ee0a"
          ],
          "question": "Does the size of the jumping shock carry the effect, or is it just the non-completion itself?",
          "mechanism": "If the lift comes from the market over-penalising any visible non-completion (the P/F/U letter), then chase non-finishers whose jumping was close to expectation should do as well as heavily shocked ones. If the shock is the cause, only the large-negative group should lift.",
          "market_miss": "Punters see the non-finish letter, not how far the run fell short of its expected Jump Index. Under the shock story, the market treats all non-finishers alike while only the badly shocked ones are one-off aberrations.",
          "prediction": {
            "direction": "positive",
            "size": "A/E 1.05 to 1.12 on 250+ bets, matching the large-shock group (the rival is supported)",
            "falsified_if": "A/E at or below 0.97 on 200+ bets while the large-shock group stays near 1.09, which would mean s carries the effect"
          },
          "rule_sketch": "code = 'Chase' AND finished = 0 AND s > -1.5"
        },
        "confounders": [
          "pulled-up vs fell/unseated mix differs by shock size",
          "price band: deeper shocks may drift to bigger prices",
          "GB/IRE mix",
          "single-season sample"
        ],
        "expected_survival": 0.3
      },
      {
        "title": "Test-mechanism: shocked chase non-finishers under 16/1 in GB",
        "family": "shock",
        "operator": "sharpen",
        "question_id": null,
        "new_question": 1,
        "parent_id": "S3-chase-not-completed-large-negative",
        "reasoning": {
          "observation_ids": [
            "O-ddd7ee0a",
            "O-69d892e9"
          ],
          "question": "Does the lift sit where the mechanism says: in horses the market still rates, in the country where non-finish edges have shown up?",
          "mechanism": "A one-off jumping shock only hides ability in a horse that has ability. Long-priced non-finishers are mostly horses that are simply poor. Previous chase cuts went 0 from 98 and 0 from 34 at 16/1 or bigger, so the effect should concentrate below 16/1. This band was chosen before the numbers, from those prior findings.",
          "market_miss": "For a fancied horse, a non-finish with a shocked Jump Index looks like lost form. The figures say it was an aberrant jumping run, which the form book records only as P or F.",
          "prediction": {
            "direction": "positive",
            "size": "A/E 1.12 to 1.25 on 120+ bets",
            "falsified_if": "A/E at or below 1.0 on 100+ bets"
          },
          "rule_sketch": "code = 'Chase' AND finished = 0 AND s <= -1.5 AND region = 'GB' AND sp_decimal < 17"
        },
        "confounders": [
          "calendar effect after April 2026 seen in sibling cuts",
          "overlap with existing chase-error non-finisher shadow rule",
          "trainer concentration",
          "small bet count"
        ],
        "expected_survival": 0.2
      }
    ],
    "findings": [
      {
        "statement": "Chase non-finishers with a large negative jumping shock were flat in discovery (1.007 on 214) and only mildly positive on the holdout (1.095 on 254, p 0.364). The holdout lift was GB-only (GB 1.117, IRE 0.971), the same split as other chase non-finisher cuts.",
        "kind": "method",
        "observation_ids": [
          "O-ddd7ee0a"
        ]
      }
    ]
  }

Timeline

1. Job 2108 gave up: OperationalError: no such column: code Traceback (most recent call last): File "/app/lab/src/gallops/orchestrator/loop

   16 Oct, 23:51 · engine:compute-1

2. compute-1 took job 2108 (forward_check, attempt 5)

   16 Oct, 23:51 · compute-1

3. Job 2108 will retry: OperationalError: no such column: code Traceback (most recent call last): File "/app/lab/src/gallops/orchestrator/loop

   16 Oct, 23:43 · engine:compute-1

4. compute-1 took job 2108 (forward_check, attempt 4)

   16 Oct, 23:43 · compute-1

5. Job 2108 will retry: OperationalError: no such column: code Traceback (most recent call last): File "/app/lab/src/gallops/orchestrator/loop

   16 Oct, 23:39 · engine:compute-1

6. compute-1 took job 2108 (forward_check, attempt 3)

   16 Oct, 23:39 · compute-1

7. Job 2108 will retry: OperationalError: no such column: code Traceback (most recent call last): File "/app/lab/src/gallops/orchestrator/loop

   16 Oct, 23:37 · engine:compute-1

8. compute-1 took job 2108 (forward_check, attempt 2)

   16 Oct, 23:37 · compute-1

9. Job 2108 will retry: OperationalError: no such column: code Traceback (most recent call last): File "/app/lab/src/gallops/orchestrator/loop

   16 Oct, 23:36 · engine:compute-1