Dose-response: the stiff-track edge is weaker at seven furlongs to a mile than in sprints

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
   • H181-proven-over-further-at-a-stiff-track (H181: proven over further at a stiff track) was GREEN in the earlier research: A/E 1.08 on 9279 bets (periods 1.06 / 1.07 / 1.12); now PROMISING. O-7cdd45d9
2. AskedDoes the edge shrink as today's trip gets longer, as the hill-adds-stamina mechanism predicts? answered
3. MechanismThe uphill finish adds a roughly fixed stamina cost, which is a larger share of a sprint than of a mile, so stamina converts into speed most in short races.
4. Why the market misses itThe market prices the trip drop the same way at every distance and does not scale it to how much the hill lengthens the race.
5. Predictedpositive, A/E 1.03 to 1.10 on 400+ bets, clearly below the sprint 1.51
6. Falsified ifA/E at or above the sprint segment's figure, or below 0.98
7. Rule sketchrace_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1760 AND track_stiff_finish = 1 AND last_win_distance_yards >= distance_yards + 330
8. question id

Mechanism and rationale

The uphill finish adds a roughly fixed stamina cost, which is a larger share of a sprint than of a mile, so stamina converts into speed most in short races.

The rule

race_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1760 AND track_stiff_finish = 1 AND last_win_yards >= distance_yards + 330

Reads: race_type, distance_yards, track_stiff_finish, last_win_yards

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

189cffe7d57ad643d9031740ed94cb1d92b8d9187be4f3a6ffc591a90f032656

Frozen

not frozen

The definition can still change.

Confounders

• Horses dropping from 1m4f or more are a different type (staying handicappers, often exposed and well handicapped on a falling mark); refinement 3 removes them
• Handicap mark: a horse whose last win came over further may now be weighted on a mark that has dropped since
• Going: soft ground at stiff tracks raises the stamina demand on its own and may carry the effect (record going_group split)
• Track identity: few stiff-finish flat tracks (e.g. Newmarket, Sandown, Beverley, Carlisle, Galway) — one course could carry the result
• Distance-band composition: 7f vs 1m races differ in field size and class mix, so the band-level A/E is not purely a trip dose
• Longshot artefact: lift concentrated at 16/1+ would not count as evidence (judge at 14/1 or shorter)
• Column choice: last_win_yards (parent H181) vs last_win_distance_yards (sprint system) should be confirmed identical before comparing to the sprint 1.51

Control

{
  "description": "Other runners in the same races should score about 1.00 or a little below (they absorb the selection's surplus). The key comparison for the mechanism is the same trip-drop at the same distances on non-stiff tracks: if the hill is what converts stamina into speed, that group should score near 1.00 and clearly below the stiff-track rule. If it matches the stiff-track rule, the edge is the trip drop (rival H-9be58fc5), not the hill. The dose-response itself is read against the sprint segment (A/E 1.51) — this band should be lower.",
  "rule": "race_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1760 AND track_stiff_finish = 0 AND last_win_yards >= distance_yards + 330"
}

Discovery result

fail broad rule · 16 Oct, 12:43

Bets

109

10 winners

A/E

0.90

95%: 0.43 to 1.65

SP return

+95.1%

level stakes

Betfair SP return

+143.1%

107 bets with a BSP

Strike rate 9.2%. p = 0.68. Same-race control: the other runners in these races had A/E 1.01, so the selection did not 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 · 41s · 16 Oct, 12:42

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1760 AND track_stiff_finish = 1 AND last_win_yards >= distance_yards + 330",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "race_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1760 AND track_stiff_finish = 1 AND last_win_yards >= distance_yards + 330 AND sp_decimal <= 15"
      },
      {
        "label": "seven furlongs only (inner dose step)",
        "rule": "race_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1540 AND track_stiff_finish = 1 AND last_win_yards >= distance_yards + 330"
      },
      {
        "label": "drop of under a mile (excludes 1m4f+ stayer types)",
        "rule": "race_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1760 AND track_stiff_finish = 1 AND last_win_yards >= distance_yards + 330 AND last_win_yards < distance_yards + 880"
      }
    ],
    "control": {
      "description": "Other runners in the same races should score about 1.00 or a little below (they absorb the selection's surplus). The key comparison for the mechanism is the same trip-drop at the same distances on non-stiff tracks: if the hill is what converts stamina into speed, that group should score near 1.00 and clearly below the stiff-track rule. If it matches the stiff-track rule, the edge is the trip drop (rival H-9be58fc5), not the hill. The dose-response itself is read against the sprint segment (A/E 1.51) — this band should be lower.",
      "rule": "race_type = 'Flat' AND distance_yards > 1340 AND distance_yards <= 1760 AND track_stiff_finish = 0 AND last_win_yards >= distance_yards + 330"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Horses dropping from 1m4f or more are a different type (staying handicappers, often exposed and well handicapped on a falling mark); refinement 3 removes them",
      "Handicap mark: a horse whose last win came over further may now be weighted on a mark that has dropped since",
      "Going: soft ground at stiff tracks raises the stamina demand on its own and may carry the effect (record going_group split)",
      "Track identity: few stiff-finish flat tracks (e.g. Newmarket, Sandown, Beverley, Carlisle, Galway) — one course could carry the result",
      "Distance-band composition: 7f vs 1m races differ in field size and class mix, so the band-level A/E is not purely a trip dose",
      "Longshot artefact: lift concentrated at 16/1+ would not count as evidence (judge at 14/1 or shorter)",
      "Column choice: last_win_yards (parent H181) vs last_win_distance_yards (sprint system) should be confirmed identical before comparing to the sprint 1.51"
    ],
    "mechanism": "An uphill finish adds a roughly fixed stamina cost to a race. In a sprint that cost is a large share of the total effort, so a horse with proven stamina (a win at least 330 yards further) turns it into a real edge late in the race. At 7f to 1m the race already tests stamina for every runner, so the extra cost of the hill is a smaller share and the proven stayer's edge should shrink. The prediction is a positive edge that is weaker than in sprints, and weaker at a mile than at 7f.",
    "market_miss": "The market discounts a horse dropping in trip by about the same amount at every distance. It does not scale that discount to how much the hill lengthens the effective race, so sprints at stiff tracks are mispriced most and 7f to 1m races only mildly.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.03 to 1.10 on 400+ bets, clearly below the sprint segment's 1.51, with 7f at or above 1m",
      "falsified_if": "A/E at or above the sprint segment's figure (1.51), or below 0.98; or the non-stiff-track control at the same distances scores the same as or higher than the rule (the effect is then the trip drop, not the hill)"
    },
    "notes": "This is a dose-response test, and its scientific value is in where it sits relative to the sprint segment. A result of about 1.03–1.05 that fails the 1.05 discovery floor would still SUPPORT the mechanism, and should go on to falsification and forward tracking rather than be read as a refutation. I used last_win_yards so the rule is an exact subset of the measured parent H181 (1.08 on 9,279). The rule sketch used last_win_distance_yards. Equivalence with that column (which the sprint system uses) is an open method finding; check it before setting this figure against 1.51. Report the 7f/1m split, the going_group split, the per-course split and the price-band split (≤14/1 vs 16/1+). min_bets is raised to 300 because the prediction asks for 400+."
  }

Timeline

1. Job 1422 finished

   16 Oct, 12:43 · engine:compute-1

2. H-4aa47f2c Discovery → KILLED fails: 109 bets, needs 300; A/E 0.896, needs 1.05; p 0.6768, needs 0.05 or lower; periods [0.88], each needs 1.0; A/E 0.896 against same-race control 1.015, needs 0.03 above (not a niche pass: A/E 0.896 below 1.05; p 0.6768085388826535 above 0.5; does not beat the same-race control 1.015; a period points the other way [0.88])

   16 Oct, 12:43 · engine:evaluator

3. discovery result for H-4aa47f2c (broad rule): fail

   16 Oct, 12:43 · engine:evaluator

4. H-4aa47f2c Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 12:43 · engine:orchestrator

5. compute-1 took job 1422 (evaluate, attempt 1)

   16 Oct, 12:43 · compute-1

6. Job 1419 finished

   16 Oct, 12:42 · engine:ai-1

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

   16 Oct, 12:42 · engine:scheduler

8. H-4aa47f2c Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 12:42 · agent:designer@v1

9. ai-1 took job 1419 (design, attempt 1)

   16 Oct, 12:42 · ai-1

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

    16 Oct, 12:41 · engine:scheduler

11. New idea H-4aa47f2c: Dose-response: the stiff-track edge is weaker at seven furlongs to a mile than in sprints (test-mechanism)

    16 Oct, 12:41 · 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.