Rival: the Irish null is the stiff flag, not Ireland; the lift needs a short uphill straight

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-cbd60237 (Generalise: any proven stamina (last run or last win 330+ yards furthe) passed discovery as a niche: A/E 1.54 on 100 bets (p 0.047), same-race control 0.91. O-c404ad04
   • H-b079598e (Rival: the stiff-sprint lift is one or two GB courses, not the hill) pointed the predicted way but failed the bar: A/E 1.75 on 15 bets (p 0.248), control 0.91. O-c593bf79
2. AskedWhy does Ireland keep showing no lift (IRE 0.74 on 25, earlier 0.85 on 23)? Is track_stiff_finish coding Irish courses whose rise comes before the sprint straight, or that have no straight sprint course at all? testing
3. MechanismThe hill only adds stamina demand if sprinters meet it at the end of a straight run. Where the stiff_finish flag marks a rise on a round course, or a track with a short straight, sprints do not test stamina in the same way. That would explain the Irish nulls without any region effect.
4. Why the market misses itThe course guide flag is the same everywhere, so form readers treat all stiff tracks alike. The pricing of trip-droppers will not tell apart a hill at the end of a 5f straight from a rise on a turning track.
5. Predictedpositive, A/E 1.20 to 1.40 on 200+ bets where the straight is at least five furlongs, with IRE runners in that set not clearly below GB
6. Falsified ifA/E at or below 1.05, or Irish long-straight stiff tracks still near 0.8 while GB holds, which would make it a region or market effect rather than track shape

Mechanism and rationale

The hill only adds stamina demand if sprinters meet it at the end of a straight run. Where the stiff_finish flag marks a rise on a round course, or a track with a short straight, sprints do not test stamina in the same way. That would explain the Irish nulls without any region effect.

The rule

race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND track_straight_furlongs >= 5 AND last_distance_yards >= distance_yards + 330 AND sp_decimal <= 15

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

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

dafbd1a796fa2a60ff7d395a05a922605d53fd3422594b13d7a55405a4c49d50

Frozen

not frozen

The definition can still change.

Confounders

• track_straight_furlongs is a course-guide figure: coarse, possibly missing (NULL drops out of the rule) or not specific to the straight-course used for a given race
• few Irish stiff tracks have a 5f+ straight, so the IRE arm may stay too thin to separate region from shape
• draw bias on straight courses (high/low side) can mimic or mask a stamina effect
• one or two GB courses (e.g. a dominant stiff straight-course venue) may carry the whole result; check per-course share
• region and straight length may be collinear, so a long-straight result that is all GB cannot by itself refute the region explanation
• the vault already shows the broad stiff-sprint edge is modest (~1.17) and close to the non-stiff mirror (1.13), so part of any lift is the general trip-drop effect

Control

{
  "description": "Same-race other runners should run near A/E 1.0 (about 0.95 to 1.0); if they also beat their prices, the lift is a property of the race or course, not of stamina. The key comparison group is the mirror at stiff tracks with a short straight (under 5 furlongs) or a missing straight figure: the mechanism predicts those trip-droppers run near 1.0, clearly below the main rule. If the short-straight mirror matches the long-straight rule, track shape does not explain the lift.",
  "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND (track_straight_furlongs < 5 OR track_straight_furlongs IS NULL) AND last_distance_yards >= distance_yards + 330 AND sp_decimal <= 15"
}

Discovery result

fail broad rule · 16 Oct, 13:23

Bets

4

1 winners

A/E

2.77

95%: 0.07 to 15.43

SP return

+200.0%

level stakes

Betfair SP return

–

4 bets with a BSP

Strike rate 25.0%. p = 0.30. Same-race control: the other runners in these races had A/E 0.83, 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 · 46s · 16 Oct, 13:20

  Output (schema-checked)

  {
    "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND track_straight_furlongs >= 5 AND last_distance_yards >= distance_yards + 330 AND sp_decimal <= 15",
    "refinements": [
      {
        "label": "race run wholly on the straight",
        "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND track_straight_furlongs >= 5 AND last_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND distance_yards <= track_straight_furlongs * 220"
      },
      {
        "label": "IRE long-straight stiff tracks only (direct test of the region rival)",
        "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND track_straight_furlongs >= 5 AND last_distance_yards >= distance_yards + 330 AND sp_decimal <= 15 AND region = 'IRE'"
      }
    ],
    "control": {
      "description": "Same-race other runners should run near A/E 1.0 (about 0.95 to 1.0); if they also beat their prices, the lift is a property of the race or course, not of stamina. The key comparison group is the mirror at stiff tracks with a short straight (under 5 furlongs) or a missing straight figure: the mechanism predicts those trip-droppers run near 1.0, clearly below the main rule. If the short-straight mirror matches the long-straight rule, track shape does not explain the lift.",
      "rule": "race_type = 'Flat' AND distance_yards <= 1340 AND track_stiff_finish = 1 AND (track_straight_furlongs < 5 OR track_straight_furlongs IS NULL) AND last_distance_yards >= distance_yards + 330 AND sp_decimal <= 15"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1.05,
      "max_p": 0.2
    },
    "confounders": [
      "track_straight_furlongs is a course-guide figure: coarse, possibly missing (NULL drops out of the rule) or not specific to the straight-course used for a given race",
      "few Irish stiff tracks have a 5f+ straight, so the IRE arm may stay too thin to separate region from shape",
      "draw bias on straight courses (high/low side) can mimic or mask a stamina effect",
      "one or two GB courses (e.g. a dominant stiff straight-course venue) may carry the whole result; check per-course share",
      "region and straight length may be collinear, so a long-straight result that is all GB cannot by itself refute the region explanation",
      "the vault already shows the broad stiff-sprint edge is modest (~1.17) and close to the non-stiff mirror (1.13), so part of any lift is the general trip-drop effect"
    ],
    "mechanism": "A stiff finish adds stamina demand to a sprint only when the runners meet the rise at the end of a straight, flat-out run. Where the stiff_finish flag marks a rise on a turning course or a track with a short straight, sprinters come off a bend, ease and re-accelerate, so the hill tests stamina less. Horses whose last run was 330+ yards further carry that stamina into the sprint and are favoured only where the straight is at least 5 furlongs. If true, this explains the Irish null through track shape (few Irish stiff tracks have a long straight), not region.",
    "market_miss": "The course-guide stiff-finish flag reads the same at every track, so form readers and the market treat all stiff tracks alike and price a trip-dropper at a long uphill straight the same as one at a stiff turning track. The run over further also looks like poor speed form for a sprint, which keeps its price up.",
    "prediction": {
      "direction": "positive",
      "size": "A/E about 1.20 to 1.40 on the long-straight set (vault likely nearer 1.15 to 1.25), with the short-straight stiff mirror near 1.0 and IRE long-straight runners not clearly below GB",
      "falsified_if": "A/E at or below 1.05 on the main rule; or the short-straight stiff mirror scores as well as the long-straight rule; or IRE long-straight stiff runners stay near 0.8 while GB holds, which would point to a region or market effect, not track shape"
    },
    "notes": "The rule sketch is sound and kept unchanged: every column is pre-off and none reads the result. Expect fewer than 150 bets, because the parent rule had 83 at discovery and this rule is narrower, so it will probably go on to falsification and forward tracking rather than pass discovery outright. The IRE refinement is a diagnostic for the region rival, not a betting rule. The wholly-on-the-straight refinement uses 220 yards per furlong. Check the per-course share of bets, and the GB/IRE split inside the main rule."
  }

Timeline

1. Job 1584 finished

   16 Oct, 13:23 · engine:compute-1

2. H-ae0702d7 Discovery → KILLED fails: 4 bets, needs 150; p 0.303, needs 0.05 or lower; periods [], each needs 1.0 (not a niche pass: only 4 bets)

   16 Oct, 13:23 · engine:evaluator

3. discovery result for H-ae0702d7 (broad rule): fail

   16 Oct, 13:23 · engine:evaluator

4. H-ae0702d7 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 13:23 · engine:orchestrator

5. compute-1 took job 1584 (evaluate, attempt 1)

   16 Oct, 13:23 · compute-1

6. Job 1582 finished

   16 Oct, 13:20 · engine:ai-1

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

   16 Oct, 13:20 · engine:scheduler

8. H-ae0702d7 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 13:20 · agent:designer@v1

9. ai-1 took job 1582 (design, attempt 1)

   16 Oct, 13:20 · ai-1

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

    16 Oct, 13:19 · engine:scheduler

11. New idea H-ae0702d7: Rival: the Irish null is the stiff flag, not Ireland; the lift needs a short uphill straight (rival-explanation)

    16 Oct, 13:19 · 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.