Rival: the stiff-sprint lift is one or two GB courses, not the hill

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-2b483ac5 (Generalise: last run over further (not just last win) pays in stiff Fl) passed discovery as a niche: A/E 1.54 on 83 bets (p 0.072), same-race control 0.92. O-2f9e434c
   • H-fce3877f (Rival: the stiff-sprint edge is a course or draw quirk, not the hill) pointed the predicted way but failed the bar: A/E 1.37 on 20 bets (p 0.336), control 0.95. O-59976f8e
2. AskedWhy do Irish stiff-track sprints show no lift (IRE 0.85 on 23) while GB shows 1.75 on 60? Is track_stiff_finish coding a different kind of finish in Ireland, or are the GB bets concentrated on one or two courses? testing
3. MechanismIf the hill causes the edge, it should hold on GB stiff tracks where the sprint is not run entirely on one long straight (straight shorter than the race trip), because the climb at the end is still there. If the lift lives only on long straight courses, a draw or course quirk is the likelier cause.
4. Why the market misses itDraw and course quirks on long straight courses are published and priced. A stamina effect that also holds on turning stiff sprints would be the part the market under-weights.
5. Predictedpositive, A/E 1.15 or more on 40+ bets on turning stiff sprints in GB
6. Falsified ifA/E at or below 1.00 on turning stiff GB sprints while straight-course sprints carry the lift, which would point to a course or draw quirk
7. Rule sketchrace_type = 'Flat' AND region = 'GB' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND sp_decimal <= 15 AND last_distance_yards >= distance_yards + 330 AND track_straight_furlongs * 220 < distance_yards

Mechanism and rationale

If the hill causes the edge, it should hold on GB stiff tracks where the sprint is not run entirely on one long straight (straight shorter than the race trip), because the climb at the end is still there. If the lift lives only on long straight courses, a draw or course quirk is the likelier cause.

The rule

race_type = 'Flat' AND region = 'GB' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND sp_decimal <= 15 AND last_distance_yards >= distance_yards + 330 AND track_straight_furlongs * 220 < distance_yards

Reads: race_type, region, distance_yards, track_stiff_finish, sp_decimal, last_distance_yards, track_straight_furlongs

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

a3b982e57bff123152c80702bdf88a0f87c02d37bc944f02d8fbfb9dbad85430

Frozen

not frozen

The definition can still change.

Confounders

• Very thin: the parent GB system had only 60 bets, so the turning subset may be under 40 and cannot reach the 150-bet floor in discovery; read direction and size only and send on to forward tracking
• track_straight_furlongs may be missing (rows drop out silently) or approximate, which can misclassify courses where the straight is close to the race trip (e.g. 6f straights against 6f races)
• Few distinct courses on each side of the split, so a result on either side may still be one course; check how wins spread by course
• Field sizes are smaller on turning tracks, which changes draw effects and win rates
• The meeting mix by season differs between courses (e.g. summer festivals on long straight courses)
• Stalls position on turning sprints mixes a draw bias in with the stamina effect

Control

{
  "description": "Two comparisons. (1) The lab's same-race control: other runners in these turning stiff GB sprints should run to about their prices (A/E near 0.90-1.00, as in the parent system's 0.90 control). If they also run well above 1.0, the lift belongs to the races or courses, not to stamina. (2) A straight-course mirror: the same rule on GB stiff sprints run wholly on the straight (straight at least as long as the race trip). The rival says the lift sits there and not on turning tracks; the hill hypothesis says both should be positive. Runners with NULL track_straight_furlongs fall out of both rules and are not counted.",
  "rule": "race_type = 'Flat' AND region = 'GB' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND sp_decimal <= 15 AND last_distance_yards >= distance_yards + 330 AND track_straight_furlongs * 220 >= distance_yards"
}

Discovery result

fail broad rule · 16 Oct, 13:15

Bets

15

3 winners

A/E

1.75

95%: 0.36 to 5.10

SP return

+65.6%

level stakes

Betfair SP return

–

14 bets with a BSP

Strike rate 20.0%. p = 0.25. Same-race control: the other runners in these races had A/E 0.91, 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 · 28s · 16 Oct, 13:12

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND region = 'GB' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND sp_decimal <= 15 AND last_distance_yards >= distance_yards + 330 AND track_straight_furlongs * 220 < distance_yards",
    "refinements": [],
    "control": {
      "description": "Two comparisons. (1) The lab's same-race control: other runners in these turning stiff GB sprints should run to about their prices (A/E near 0.90-1.00, as in the parent system's 0.90 control). If they also run well above 1.0, the lift belongs to the races or courses, not to stamina. (2) A straight-course mirror: the same rule on GB stiff sprints run wholly on the straight (straight at least as long as the race trip). The rival says the lift sits there and not on turning tracks; the hill hypothesis says both should be positive. Runners with NULL track_straight_furlongs fall out of both rules and are not counted.",
      "rule": "race_type = 'Flat' AND region = 'GB' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND sp_decimal <= 15 AND last_distance_yards >= distance_yards + 330 AND track_straight_furlongs * 220 >= distance_yards"
    },
    "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": [
      "Very thin: the parent GB system had only 60 bets, so the turning subset may be under 40 and cannot reach the 150-bet floor in discovery; read direction and size only and send on to forward tracking",
      "track_straight_furlongs may be missing (rows drop out silently) or approximate, which can misclassify courses where the straight is close to the race trip (e.g. 6f straights against 6f races)",
      "Few distinct courses on each side of the split, so a result on either side may still be one course; check how wins spread by course",
      "Field sizes are smaller on turning tracks, which changes draw effects and win rates",
      "The meeting mix by season differs between courses (e.g. summer festivals on long straight courses)",
      "Stalls position on turning sprints mixes a draw bias in with the stamina effect"
    ],
    "mechanism": "A horse whose last run was at least 330 yards further has shown it can still be galloping late. An uphill finish rewards exactly that in a sprint. If the hill causes the edge, it must show on GB stiff tracks where the sprint includes a bend (straight shorter than the race trip), because the climb at the end is still there. If the lift lives only on sprints run wholly on one long straight, a draw or course quirk is the likelier cause.",
    "market_miss": "Draw and course quirks on long straight sprint courses are published and priced. Bettors read a drop in trip to sprint distance as a lack of speed, and on turning tracks they put a sprinter's tactical speed ahead of stamina. A stamina edge that also holds on turning stiff sprints would be the part the market under-weights.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.15 or more on the turning stiff GB sprints, likely on 40 or more bets, near the parent's 1.75 GB figure but lower",
      "falsified_if": "A/E at or below 1.00 on turning stiff GB sprints while the straight-course mirror carries the lift (A/E 1.15 or more). That would point to a course or draw quirk rather than the hill."
    },
    "notes": "I kept the rule sketch unchanged: all its columns are pre-off, and track_straight_furlongs * 220 turns furlongs into yards. No refinements: the price cap and the GB restriction are already in the broad rule, and the sample is too thin to slice further. Expect fewer than 150 bets, so discovery will judge the direction rather than significance. Report the turning and straight results side by side, with the bets for each course, so the rival can be judged."
  }

Timeline

1. Job 1552 finished

   16 Oct, 13:15 · engine:compute-1

2. H-b079598e Discovery → KILLED fails: 15 bets, needs 150; p 0.2478, needs 0.05 or lower; periods [], each needs 1.0 (not a niche pass: only 15 bets)

   16 Oct, 13:15 · engine:evaluator

3. discovery result for H-b079598e (broad rule): fail

   16 Oct, 13:15 · engine:evaluator

4. H-b079598e Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 13:15 · engine:orchestrator

5. compute-1 took job 1552 (evaluate, attempt 1)

   16 Oct, 13:15 · compute-1

6. Job 1550 finished

   16 Oct, 13:12 · engine:ai-1

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

   16 Oct, 13:12 · engine:scheduler

8. H-b079598e Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 13:12 · agent:designer@v1

9. ai-1 took job 1550 (design, attempt 1)

   16 Oct, 13:12 · ai-1

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

    16 Oct, 13:11 · engine:scheduler

11. New idea H-b079598e: Rival: the stiff-sprint lift is one or two GB courses, not the hill (rival-explanation)

    16 Oct, 13:11 · 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.