Replicate: hurdle winners lightly raced on the Flat in British summer staying Flat handicaps

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-cf4b198a (Sharpen: proven hurdlers lightly raced on the Flat in Irish summer sta) passed discovery as a niche: A/E 1.26 on 50 bets (p 0.341), same-race control 0.94. O-d3f516ff
2. AskedIs the lift strongest where the Flat mark has the least information, i.e. horses lightly raced on the Flat but proven over hurdles? answered
3. MechanismIf the Flat mark really lags behind hurdle ability, British handicappers should face the same gap with jumpers sent Flat handicapping in summer.
4. Why the market misses itBritish punters also read Flat form within the Flat code, and the hurdle win sits outside the Flat form lines.
5. Predictedpositive, A/E 1.05 to 1.12 on 150+ bets
6. Falsified ifA/E at or below 1.00 on 120+ bets, which would make the effect Irish-specific (yard or meeting) rather than a property of the mark
7. Rule sketchregion = 'GB' AND race_type = 'Flat' AND is_handicap = 1 AND month BETWEEN 6 AND 9 AND distance_yards >= 2640 AND hurdle_wins >= 1 AND flat_runs <= 6
8. question id

Mechanism and rationale

If the Flat mark really lags behind hurdle ability, British handicappers should face the same gap with jumpers sent Flat handicapping in summer.

The rule

region = 'GB' AND race_type = 'Flat' AND is_handicap = 1 AND month BETWEEN 6 AND 9 AND distance_yards >= 2640 AND hurdle_wins >= 1 AND flat_runs <= 6

Reads: region, race_type, is_handicap, month, distance_yards, hurdle_wins, flat_runs

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

22d849524bb54c3dc7ec2a01afec08844765434baec2530b8142cea4b7c597e1

Frozen

not frozen

The definition can still change.

Confounders

• The GB handicapper may already fold hurdle form into Flat marks more fully than the Irish handicapper, which would close the gap for policy reasons, not because the mechanism is wrong
• GB staying handicaps have more Flat-only stayers, so a dual-purpose horse on a lenient mark may still face stronger opposition
• A few dual-purpose yards (e.g. specialist trainers who run jumpers in summer staying handicaps) could carry the sample; check how the winners are spread across trainers
• Price profile: the Irish lift came from short-priced winners (none at 16/1 or bigger). A GB lift that rests on a few big-price winners would not be the same effect
• hurdle_wins is only available from 2020-10-01, so runs earlier in the window are left out, and a horse's pre-2020 hurdle wins may be undercounted, which would put some real hurdle winners in the non-selected group
• Course mix: GB summer staying handicaps are concentrated at a few tracks and meetings (e.g. Goodwood, York), so a course effect could stand in for the mark effect

Control

Discovery result

fail broad rule · 17 Oct, 03:29

Bets

48

7 winners

A/E

0.99

95%: 0.40 to 2.05

SP return

-40.4%

level stakes

Betfair SP return

-29.0%

48 bets with a BSP

Strike rate 14.6%. p = 0.56. Same-race control: the other runners in these races had A/E 1.00, 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 · 44s · 17 Oct, 03:24

  Output (schema-checked)

  {
    "rule": "region = 'GB' AND race_type = 'Flat' AND is_handicap = 1 AND month BETWEEN 6 AND 9 AND distance_yards >= 2640 AND hurdle_wins >= 1 AND flat_runs <= 6",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "region = 'GB' AND race_type = 'Flat' AND is_handicap = 1 AND month BETWEEN 6 AND 9 AND distance_yards >= 2640 AND hurdle_wins >= 1 AND flat_runs <= 6 AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "Same-race controls: the other runners in the same GB summer staying Flat handicaps. If the information gap is real, the selections should beat their prices while their race-mates sit near or slightly below 1.00 (the Irish parent showed 1.26 against a control of 0.94). If the race-mates also come in clearly above 1.00, the lift belongs to the race type (summer staying handicaps being mispriced generally), not to the hurdle-winner gap. The lab's built-in same-race comparison is enough, so no separate control rule is needed.",
      "rule": null
    },
    "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": [
      "The GB handicapper may already fold hurdle form into Flat marks more fully than the Irish handicapper, which would close the gap for policy reasons, not because the mechanism is wrong",
      "GB staying handicaps have more Flat-only stayers, so a dual-purpose horse on a lenient mark may still face stronger opposition",
      "A few dual-purpose yards (e.g. specialist trainers who run jumpers in summer staying handicaps) could carry the sample; check how the winners are spread across trainers",
      "Price profile: the Irish lift came from short-priced winners (none at 16/1 or bigger). A GB lift that rests on a few big-price winners would not be the same effect",
      "hurdle_wins is only available from 2020-10-01, so runs earlier in the window are left out, and a horse's pre-2020 hurdle wins may be undercounted, which would put some real hurdle winners in the non-selected group",
      "Course mix: GB summer staying handicaps are concentrated at a few tracks and meetings (e.g. Goodwood, York), so a course effect could stand in for the mark effect"
    ],
    "mechanism": "A horse that has won over hurdles but has run six or fewer times on the Flat gets a Flat handicap mark from thin Flat evidence. That mark can lag behind the stamina and ability the hurdle win shows. If the lag is a general property of how Flat marks are set, and not something specific to Irish yards or festival meetings, British summer staying Flat handicaps (2m+, June to September) should show the same edge.",
    "market_miss": "British Flat punters and odds-compilers read form within the Flat code: Flat runs, Flat ratings and Flat trainer patterns. A hurdle win sits outside those form lines, so a lightly raced Flat profile looks exposed or modest and gets priced on its Flat figures. The few Flat runs also give the market little to correct with.",
    "prediction": {
      "direction": "positive",
      "size": "A/E about 1.05 to 1.12 on 150+ bets, with same-race controls near 1.00 or below, and the lift concentrated at 14/1 or shorter",
      "falsified_if": "A/E at or below 1.00 on 120+ bets, which would make the Irish effect Irish-specific (yard or meeting) rather than a property of the mark. Also falsified if the selections do not beat their same-race controls"
    },
    "notes": "This is a straight replication of H-cf4b198a (IRE 1.26 on 50, control 0.94), with only the region changed so the test stays clean. All columns are pre-off and available in GB (hurdle_wins from 2020-10-01, flat_runs and month from 2020-07-01). The only refinement is a price cap at 14/1 or shorter, because the Irish winners all came at short prices. If GB is thin (under 120 bets), report it as inconclusive, not as falsified."
  }

Timeline

1. Job 2666 finished

   17 Oct, 03:29 · engine:compute-1

2. H-a11f4700 Discovery → KILLED fails: 48 bets, needs 150; A/E 0.994, needs 1.05; p 0.5564, needs 0.05 or lower; periods [0.99], each needs 1.0; A/E 0.994 against same-race control 0.998, needs 0.03 above (not a niche pass: A/E 0.994 below 1.05; p 0.5563799939738971 above 0.5; does not beat the same-race control 0.998)

   17 Oct, 03:29 · engine:evaluator

3. discovery result for H-a11f4700 (broad rule): fail

   17 Oct, 03:29 · engine:evaluator

4. H-a11f4700 Designed → RESEARCHdiscovery measuring on the discovery zone

   17 Oct, 03:29 · engine:orchestrator

5. compute-1 took job 2666 (evaluate, attempt 1)

   17 Oct, 03:29 · compute-1

6. Job 2663 finished

   17 Oct, 03:25 · engine:ai-1

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

   17 Oct, 03:25 · engine:scheduler

8. H-a11f4700 Proposed → RESEARCHdesigned pre-spec written and hashed

   17 Oct, 03:25 · agent:designer@v1

9. ai-1 took job 2663 (design, attempt 1)

   17 Oct, 03:24 · ai-1

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

    17 Oct, 03:22 · engine:scheduler

11. New idea H-a11f4700: Replicate: hurdle winners lightly raced on the Flat in British summer staying Flat handicaps (replicate)

    17 Oct, 03:22 · 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.