Test-mechanism: held-up chase pull-ups dropped back to hurdles

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-fbabeb50 (Sharpen: held-up hurdle pull-ups sent chasing next time) passed discovery as a niche: A/E 1.5 on 76 bets (p 0.17), same-race control 0.96. O-2b845062
   • H-f0aa49df (Rival: any pull-up switching code beats its price, held-up or not) was predicted positive but came out the other way: A/E 0.87 on 161 bets (p 0.414 for the opposite direction). O-cd625e74
   • H-fbabeb50 (Sharpen: held-up hurdle pull-ups sent chasing next time) survived falsification: 4 of 4 attacks left it standing; it is frozen for its one vault query. O-f1f004ec
2. AskedWithin the 226 jumps-to-jumps switchers, is the lift in hurdle-to-chase (a first try over fences) or chase-to-hurdle (fences taken away), and does a jumping error in the pull-up matter? answered
3. MechanismIf the cause is a pull-up that hides ability (the horse was looked after, never asked a question) and today's code change resets the market's reading, then chase-to-hurdle switchers should lift as well. If the cause is new conditions suiting the horse (a chasing type finally over fences), the lift should be weak or absent here.
4. Why the market misses itThe P in the form figures and a code change both push the price out. A chaser dropped back to hurdles after pulling up looks like a horse in decline, not one that was simply looked after.
5. Predictedpositive, A/E 1.20 to 1.40 on 120+ bets if the hidden-ability mechanism holds

Mechanism and rationale

If the cause is a pull-up that hides ability (the horse was looked after, never asked a question) and today's code change resets the market's reading, then chase-to-hurdle switchers should lift as well. If the cause is new conditions suiting the horse (a chasing type finally over fences), the lift should be weak or absent here.

The rule

last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Chase' AND race_type = 'Hurdle'

Reads: last_pulled_up, last_run_style, last_race_type, race_type

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

a46247aa2caabc6ce83767a8674ff1ef0ae4d37dd32169aa1f06e84b2cb409a7

Frozen

17 Oct, 02:43

Nothing in the definition can change after this.

Confounders

• Older, declining chasers dropped back to hurdles because they no longer jump or stay. These are genuinely worse horses, which would push A/E down (refinement: age 9 or under).
• Handicap marks differ between codes. A chaser can come back on a softer or harsher hurdle mark, so a lift could come from the mark and not from hidden form (refinement: handicap to handicap).
• Race mix: the sibling hurdle-to-chase cuts lifted only after 27 April 2026 (summer jumps). Check that the result holds in both periods and is not one season's mix.
• Thin sample: chase-to-hurdle switches after a held-up pull-up are uncommon. A few big-priced winners could carry the A/E, so check the 14/1-or-shorter cut and how many winners there are.
• Long prices: earlier findings show every loss beyond par sitting at 16/1 or bigger, so a broad-rule result could be diluted or inflated by the tail.
• A pull-up for an injury or physical problem (bled, lame) is not a horse that was looked after. The data can't separate these, so they add noise against the mechanism.

Discovery result

pass broad rule (niche) · 17 Oct, 02:35

Bets

150

12 winners

A/E

1.38

95%: 0.71 to 2.40

SP return

-36.9%

level stakes

Betfair SP return

-18.9%

150 bets with a BSP

Strike rate 8.0%. p = 0.17. Same-race control: the other runners in these races had A/E 0.98, 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. 4 of 5 survived.

• control: Control: held-up chase pull-ups that stay over fencespass

  404 bets · A/E 1.08 (0.73 to 1.55) · control 1.00 · 17 Oct, 02:43

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-272871.09

  after 2026-04-271171.08

• control: Control: prominent or front-running chase pull-ups dropped to hurdlesfail

  76 bets · A/E 0.85 (0.23 to 2.17) · control 1.01 · 17 Oct, 02:43

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': 1.0, 'max_p': 0.2}info

  0 bets · · 17 Oct, 02:58

  Breakdown

  betsA/E0.71.01.3

  forward0–

Other experiments

• neighbour: drop: last_pulled_up = 1fail

  752 bets · A/E 0.93 (0.70 to 1.22) · control 1.02 · 17 Oct, 02:35

  Breakdown

  betsA/E0.71.01.3

  2025-10-01 to 2026-04-273930.90

  after 2026-04-273590.97

• neighbour: drop: last_run_style IN ('held_up', 'midfield')pass

  233 bets · A/E 1.23 (0.71 to 1.96) · control 0.99 · 17 Oct, 02:35

  Breakdown

Lineage

Agent runs

• designer v1ok · 28s · 17 Oct, 02:31

  Output (schema-checked)

  {
    "rule": "last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Chase' AND race_type = 'Hurdle'",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Chase' AND race_type = 'Hurdle' AND sp_decimal <= 15"
      },
      {
        "label": "Not an old declining chaser (age 9 or under)",
        "rule": "last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Chase' AND race_type = 'Hurdle' AND age <= 9"
      },
      {
        "label": "Handicap to handicap (mark comparable across the switch)",
        "rule": "last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Chase' AND race_type = 'Hurdle' AND is_handicap = 1 AND last_is_handicap = 1"
      }
    ],
    "control": {
      "description": "The same-race control (other runners in the same hurdle races) should sit near A/E 1.0, which shows the race type or race mix is not lifting everyone. The second comparison separates the code switch from the pull-up itself: held-up or midfield chase pull-ups that stay over fences. If the hidden-ability-plus-reset mechanism is right, the switchers should beat this group. If both groups show the same lift, the effect comes from the held-up pull-up and not from the switch.",
      "rule": "last_pulled_up = 1 AND last_run_style IN ('held_up', 'midfield') AND last_race_type = 'Chase' AND race_type = '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": [
      "Older, declining chasers dropped back to hurdles because they no longer jump or stay. These are genuinely worse horses, which would push A/E down (refinement: age 9 or under).",
      "Handicap marks differ between codes. A chaser can come back on a softer or harsher hurdle mark, so a lift could come from the mark and not from hidden form (refinement: handicap to handicap).",
      "Race mix: the sibling hurdle-to-chase cuts lifted only after 27 April 2026 (summer jumps). Check that the result holds in both periods and is not one season's mix.",
      "Thin sample: chase-to-hurdle switches after a held-up pull-up are uncommon. A few big-priced winners could carry the A/E, so check the 14/1-or-shorter cut and how many winners there are.",
      "Long prices: earlier findings show every loss beyond par sitting at 16/1 or bigger, so a broad-rule result could be diluted or inflated by the tail.",
      "A pull-up for an injury or physical problem (bled, lame) is not a horse that was looked after. The data can't separate these, so they add noise against the mechanism."
    ],
    "mechanism": "A held-up or midfield pull-up often means the horse was looked after once its chance had gone, and was never asked a question, so it hides ability. If that hidden ability, together with the market resetting its reading at a code change, causes the established hurdle-to-chase switch lift (A/E 1.50 on 76), then the same lift should show when the switch runs the other way, from fences back to hurdles. If the hurdle-to-chase lift came instead from new conditions suiting a chasing type, this reverse switch should show little or no lift.",
    "market_miss": "The P in the form figures and a drop from fences back to hurdles both push the price out. Punters read a chaser sent back over hurdles after pulling up as a horse in decline, not as one that was simply looked after. They price the bad-looking form line without telling a held-up, no-question pull-up apart from a genuine collapse.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.20 to 1.40 on 120+ bets if the hidden-ability mechanism holds",
      "falsified_if": "A/E below 1.05, or not above both the same-race control and the stay-over-fences control. Either would place the switch lift only in the hurdle-to-chase direction, which points to new conditions suiting a chaser rather than misleading form."
    },
    "notes": "I kept the rule sketch unchanged. All four columns are pre-off and readable from the discovery zone, and it is the minimal test of the mechanism. This is a test of the mechanism behind H-fbabeb50 (hurdle-to-chase held-up pull-ups, A/E 1.50 on 76), so 'midfield' stays in to match the original cut. Expect a thin sample. If the result points the right way but comes in under 150 bets, it goes to falsification and forward tracking rather than being killed."
  }

Timeline

1. Job 2540 finished

   17 Oct, 02:58 · engine:compute-1

2. forward result for H-7beeabeb (forward vs bar {'min_bets': 100, 'min_ae': 1.0, 'max_p': 0.2}): info

   17 Oct, 02:58 · engine:evaluator

3. compute-1 took job 2540 (forward_check, attempt 1)

   17 Oct, 02:58 · compute-1

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

   17 Oct, 02:58 · engine:scheduler

5. Job 2511 finished

   17 Oct, 02:43 · engine:compute-1

6. H-7beeabeb Frozen → SHADOW vault skipped: last_pulled_up, last_run_style do not exist before the discovery zone; track it forward instead

   17 Oct, 02:43 · engine:orchestrator

7. compute-1 took job 2511 (vault, attempt 1)

   17 Oct, 02:43 · compute-1

8. Job 2494 finished

   17 Oct, 02:43 · engine:compute-1

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

   17 Oct, 02:43 · engine:scheduler

10. H-7beeabeb Falsification → FROZEN survived falsification

    17 Oct, 02:43 · engine:falsification

11. falsification result for H-7beeabeb (partition: Partition: age 9 or under, priced 14/1 or shorter): pass

    17 Oct, 02:43 · engine:evaluator

12. compute-1 took job 2494 (attack, attempt 1)

    17 Oct, 02:43 · compute-1

13. Job 2493 finished

    17 Oct, 02:43 · engine:compute-1

14. falsification result for H-7beeabeb (partition: Partition: hurdle handicaps only): pass

    17 Oct, 02:43 · engine:evaluator

15. compute-1 took job 2493 (attack, attempt 1)

    17 Oct, 02:43 · compute-1

16. Job 2492 finished

    17 Oct, 02:43 · engine:compute-1

17. falsification result for H-7beeabeb (partition: Partition: held up only (midfield removed)): pass

    17 Oct, 02:43 · engine:evaluator

18. compute-1 took job 2492 (attack, attempt 1)

    17 Oct, 02:43 · compute-1

19. Job 2491 finished

    17 Oct, 02:43 · engine:compute-1

20. falsification result for H-7beeabeb (control: Control: prominent or front-running chase pull-ups dropped to hurdles): fail

    17 Oct, 02:43 · engine:evaluator