Rival: the stiff-sprint edge is a course or draw quirk, 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-4e7be756 (Rival: the stiff-sprint system's lift is a 16/1+ longshot effect) met the sealed vault supportive: A/E 1.64 on 141 bets (p 0.0068, threshold 0.0029). O-7a52c39b
   • H-4e7be756 (Rival: the stiff-sprint system's lift is a 16/1+ longshot effect) passed discovery as a niche: A/E 1.34 on 31 bets (p 0.29), same-race control 0.95. O-9ebcf76a
2. AskedIs the price-capped stiff-sprint edge (1.64 on 141 in the vault) spread across the stiff tracks, or carried by one or two courses or a draw bias at those courses? testing
3. MechanismIf the hill causes the edge, it should hold whatever the stall. If the edge comes from draw bias on straight stiff sprint courses, it should vanish once runners are confined to stalls in the bigger half of the draw number range, where the published draw biases tend to sit.
4. Why the market misses itCourse-guide draw biases are well known and priced. A stamina edge that survives across stalls would be the unpriced part.
5. Predictedpositive, A/E 1.10+ on 60+ bets in each draw half
6. Falsified ifA/E at or below 1.00 in either draw half while the other carries the lift, which would mark a draw quirk
7. Rule sketchrace_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND last_win_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND draw * 2 > field_size

Mechanism and rationale

If the hill causes the edge, it should hold whatever the stall. If the edge comes from draw bias on straight stiff sprint courses, it should vanish once runners are confined to stalls in the bigger half of the draw number range, where the published draw biases tend to sit.

The rule

race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND last_win_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND draw * 2 > field_size

Reads: race_type, distance_yards, track_stiff_finish, last_win_distance_yards, sp_decimal, draw, field_size

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

b8ece84d98e758ccf5afcb57159bc10ea69b0ef3e7edb9cfa9733641f76eef84

Frozen

not frozen

The definition can still change.

Confounders

• Small samples: the whole price-capped system had about 141 vault bets, so each draw half may hold only 60-70 bets and one or two winners can swing A/E
• Non-runners: draw is the stall number and field_size counts runners, so after withdrawals a stall can exceed field_size and the effective half shifts
• Many stiff tracks have no strong draw bias, or the bias flips with going or stalls position, which dilutes a fixed high/low split
• Field size: in small fields the draw halves barely differ, so they should behave alike whatever the cause
• Course concentration: one or two courses (for example a straight stiff 5-6f course) may supply most bets in one half and pass off a course quirk as a draw effect
• Price mix: if the halves differ in price distribution, the A/E gap may reflect price bands rather than the stall

Control

{
  "description": "The key comparison is the other draw half of the same system: price-capped stiff-sprint stayers drawn in the lower half (draw * 2 <= field_size). If the hill causes the edge, both halves should show a similar lift, around 1.10 or more, and neither should sit at or below 1.00. If one half carries the whole lift, it is a draw or course quirk. The usual same-race control (other runners in these races) should sit near 1.00. That shows the races themselves are not over-rewarding high-drawn runners in general. If the same-race high-drawn runners also show a lift, the draw explains it.",
  "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND last_win_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND draw * 2 <= field_size"
}

Discovery result

fail broad rule · 16 Oct, 13:07

Bets

20

4 winners

A/E

1.37

95%: 0.37 to 3.50

SP return

+6.2%

level stakes

Betfair SP return

–

19 bets with a BSP

Strike rate 20.0%. p = 0.34. Same-race control: the other runners in these races had A/E 0.95, 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 · 1m · 16 Oct, 13:05

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND last_win_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND draw * 2 > field_size",
    "refinements": [
      {
        "label": "fields of 8+ runners, where a draw half is a meaningful split",
        "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND last_win_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND draw * 2 > field_size AND field_size >= 8"
      }
    ],
    "control": {
      "description": "The key comparison is the other draw half of the same system: price-capped stiff-sprint stayers drawn in the lower half (draw * 2 <= field_size). If the hill causes the edge, both halves should show a similar lift, around 1.10 or more, and neither should sit at or below 1.00. If one half carries the whole lift, it is a draw or course quirk. The usual same-race control (other runners in these races) should sit near 1.00. That shows the races themselves are not over-rewarding high-drawn runners in general. If the same-race high-drawn runners also show a lift, the draw explains it.",
      "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND last_win_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND draw * 2 <= field_size"
    },
    "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": [
      "Small samples: the whole price-capped system had about 141 vault bets, so each draw half may hold only 60-70 bets and one or two winners can swing A/E",
      "Non-runners: draw is the stall number and field_size counts runners, so after withdrawals a stall can exceed field_size and the effective half shifts",
      "Many stiff tracks have no strong draw bias, or the bias flips with going or stalls position, which dilutes a fixed high/low split",
      "Field size: in small fields the draw halves barely differ, so they should behave alike whatever the cause",
      "Course concentration: one or two courses (for example a straight stiff 5-6f course) may supply most bets in one half and pass off a course quirk as a draw effect",
      "Price mix: if the halves differ in price distribution, the A/E gap may reflect price bands rather than the stall"
    ],
    "mechanism": "A stiff uphill finish rewards proven stamina in a sprint: a horse that has already won over 330+ yards further keeps going up the hill while pure sprinters fade. That effect comes from the horse and the track, not the stall, so it should appear in both halves of the draw. The rival explanation is that the lift comes from published draw biases on straight stiff sprint courses, which would put it all in one draw half.",
    "market_miss": "Course-guide draw biases are widely published and priced into SPs, so a draw-driven lift should not survive at 14/1 or shorter. A stamina-on-the-hill edge that holds in both draw halves is the part the market under-weights: punters rate sprinters on recent sprint form and mark down horses whose winning form is over further.",
    "prediction": {
      "direction": "positive",
      "size": "A/E about 1.10 or more in the high draw half (this rule), similar to the low-half control, on roughly 60+ bets each",
      "falsified_if": "A/E at or below 1.00 in this high-draw half while the low-half control carries the lift (or the reverse: this half 1.10+ while the low half is at or below 1.00), which would mark a draw or course quirk rather than the hill"
    },
    "notes": "I kept the rule sketch as written. I checked the columns with query_catalog: draw (INT, stall), field_size, and last_win_distance_yards (INT, available from 2020-09-01) all exist and are pre_off. This rule measures the high half, and the low half is the control, so the test is the comparison between them, not this half alone. Each half will probably fall short of the 150-bet floor, so expect a thin result that goes on to falsification and forward tracking. Judge it by how close the two halves are, not by the p-value alone. I did not add a track_draw_bias refinement because its text values are not documented. The field_size >= 8 refinement deals with the small-field confounder. Earlier flag finding: no stiff Flat sprints also carry track_testing_ground = 1, so that flag is not used here."
  }

Timeline

1. Job 1533 finished

   16 Oct, 13:07 · engine:compute-1

2. H-fce3877f Discovery → KILLED fails: 20 bets, needs 150; p 0.336, needs 0.05 or lower; periods [], each needs 1.0 (not a niche pass: only 20 bets)

   16 Oct, 13:07 · engine:evaluator

3. discovery result for H-fce3877f (broad rule): fail

   16 Oct, 13:07 · engine:evaluator

4. H-fce3877f Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 13:07 · engine:orchestrator

5. compute-1 took job 1533 (evaluate, attempt 1)

   16 Oct, 13:07 · compute-1

6. Job 1531 finished

   16 Oct, 13:06 · engine:ai-1

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

   16 Oct, 13:06 · engine:scheduler

8. H-fce3877f Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 13:06 · agent:designer@v1

9. ai-1 took job 1531 (design, attempt 1)

   16 Oct, 13:05 · ai-1

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

    16 Oct, 13:05 · engine:scheduler

11. New idea H-fce3877f: Rival: the stiff-sprint edge is a course or draw quirk, not the hill (rival-explanation)

    16 Oct, 13:05 · 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.