Replicate: fell-going-well chase edge in IRE

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-3cc36f8a (Rival: the 8/1 to 14/1 chase lift is generic to well-beaten chasers wi): the country 'GB' carries the effect (A/E 1.35) while the rest score IRE 0.97. O-3f2543b9
   • H85-fell-when-going-well (H85: fell when going well) was GREEN in the earlier research: A/E 1.13 on 405 bets (periods 1.15 / 1.11); now PROMISING. O-add18290
2. AskedIs the fell-when-going-well chase edge GB-led, as the chase-error edge is? answered
3. MechanismIf the edge is a misread of the form figures, it should appear wherever punters read F and U beside a horse's name, so in IRE as well as GB.
4. Why the market misses itIrish form figures show the same F or U. But Irish chase markets are dominated by big yards and may price their runners differently.
5. Predictedpositive, A/E 1.10 to 1.25 on 60+ bets
6. Falsified ifAn IRE A/E at or below 1.0 would mean the system should be restricted to GB.
7. Rule sketchregion = 'IRE' AND race_type = 'Chase' AND last_fell_going_well = 1

Mechanism and rationale

If the edge is a misread of the form figures, it should appear wherever punters read F and U beside a horse's name, so in IRE as well as GB.

The rule

region = 'IRE' AND race_type = 'Chase' AND last_fell_going_well = 1

Reads: region, race_type, last_fell_going_well

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

64e7ca941e410b4754d2c56e2be4806e7bd0b7084e12de5ff3a6027c875868b3

Frozen

not frozen

The definition can still change.

Confounders

• Small IRE sample: last_fell_going_well only exists from 2025-11-01, so the discovery pool is likely well under 150 bets, and a single result will be noisy.
• Festival concentration: Irish chase fields swell at Leopardstown at Christmas, the Dublin Racing Festival, Fairyhouse at Easter and Punchestown, where markets are sharper and fields bigger. A few winners at those meetings could decide the result.
• Dominant yards: a handful of big Irish stables dominate. Their fallers may be re-routed or priced on stable reputation rather than form figures, which could swamp any form-figure misread.
• Calendar skew: other lifts in this lab sit after 27 April 2026, so a lift that shows up in only one period should be read as possible drift.
• Overlap with the chase-error thread: IRE non-finishers there scored 0.43 on 23. Last time's fallers are by definition last-time non-finishers, so the prior points against an IRE edge.
• Price band: IRE chase-error runners went 0 from 64 at 16/1 or bigger, so a few big-price losers can drag A/E down even if shorter-priced runners hold the edge.

Discovery result

fail broad rule · 17 Oct, 02:11

Bets

53

7 winners

A/E

0.92

95%: 0.37 to 1.90

SP return

+100.3%

level stakes

Betfair SP return

+261.1%

53 bets with a BSP

Strike rate 13.2%. p = 0.63. 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 · 46s · 17 Oct, 02:09

  Output (schema-checked)

  {
    "rule": "region = 'IRE' AND race_type = 'Chase' AND last_fell_going_well = 1",
    "refinements": [
      {
        "label": "IRE, 14/1 or shorter",
        "rule": "region = 'IRE' AND race_type = 'Chase' AND last_fell_going_well = 1 AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "The lab's same-race control: other runners in the same IRE chases should sit near A/E 1.0 (about 0.95 to 1.0 after the overround). That shows any lift belongs to the fell-going-well horses and not to the races they run in. Second comparison: the GB version of the same rule, which scored about 1.30 on 165 in discovery. Comparing the two answers the question directly. If IRE is at or below 1.0 while GB holds, the system is GB-only.",
      "rule": "region = 'GB' AND race_type = 'Chase' AND last_fell_going_well = 1"
    },
    "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": [
      "Small IRE sample: last_fell_going_well only exists from 2025-11-01, so the discovery pool is likely well under 150 bets, and a single result will be noisy.",
      "Festival concentration: Irish chase fields swell at Leopardstown at Christmas, the Dublin Racing Festival, Fairyhouse at Easter and Punchestown, where markets are sharper and fields bigger. A few winners at those meetings could decide the result.",
      "Dominant yards: a handful of big Irish stables dominate. Their fallers may be re-routed or priced on stable reputation rather than form figures, which could swamp any form-figure misread.",
      "Calendar skew: other lifts in this lab sit after 27 April 2026, so a lift that shows up in only one period should be read as possible drift.",
      "Overlap with the chase-error thread: IRE non-finishers there scored 0.43 on 23. Last time's fallers are by definition last-time non-finishers, so the prior points against an IRE edge.",
      "Price band: IRE chase-error runners went 0 from 64 at 16/1 or bigger, so a few big-price losers can drag A/E down even if shorter-priced runners hold the edge."
    ],
    "mechanism": "Punters read the F or U in the form figures as a sign of poor jumping or poor form. But when the horse was going well at the time it fell, the run before the error showed ability the figures hide. If the edge is a misread of the form figures, rather than something about British markets, it should appear wherever F and U are printed beside a horse's name, so in Irish chases as well as British ones.",
    "market_miss": "Irish form figures show the same F or U, so the same surface misread is available. But Irish chase markets are dominated by big yards and well-informed money, which may price stable reputation and the trainer's intentions ahead of the form figures. That could correct the misread in Ireland but not in Britain.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.10 to 1.25, likely on 40 to 100 bets given the column only starts in November 2025",
      "falsified_if": "An IRE A/E at or below 1.0 on the broad rule falsifies the replication, and the fell-going-well chase system should be restricted to GB. An A/E above 1.0 that only comes from a 16/1+ outlier, or only from the late period, does not count as replication."
    },
    "notes": "The rule sketch is sound: every column is pre-off, and nothing reads the result. last_fell_going_well is available in IRE from 2025-11-01 only, so the broad rule will almost certainly fall short of the 150-bet floor. Under lab policy a thin result pointing the right way goes on to falsification and forward tracking, not to the bin. The one refinement, 14/1 or shorter (sp_decimal <= 15), comes from the finding that IRE chase-error runners went 0 from 64 at 16/1+. The forward bar's A/E is raised to 1.05 so that a forward record at par does not count as replication."
  }

Timeline

1. Job 2419 finished

   17 Oct, 02:11 · engine:compute-1

2. H-2df93a3d Discovery → KILLED fails: 53 bets, needs 150; A/E 0.923, needs 1.05; p 0.6336, needs 0.05 or lower; periods [], each needs 1.0; A/E 0.923 against same-race control 1.012, needs 0.03 above (not a niche pass: A/E 0.923 below 1.05; p 0.6336353379039796 above 0.5; does not beat the same-race control 1.012)

   17 Oct, 02:11 · engine:evaluator

3. discovery result for H-2df93a3d (broad rule): fail

   17 Oct, 02:11 · engine:evaluator

4. H-2df93a3d Designed → RESEARCHdiscovery measuring on the discovery zone

   17 Oct, 02:11 · engine:orchestrator

5. compute-1 took job 2419 (evaluate, attempt 1)

   17 Oct, 02:11 · compute-1

6. Job 2416 finished

   17 Oct, 02:09 · engine:ai-1

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

   17 Oct, 02:09 · engine:scheduler

8. H-2df93a3d Proposed → RESEARCHdesigned pre-spec written and hashed

   17 Oct, 02:09 · agent:designer@v1

9. ai-1 took job 2416 (design, attempt 1)

   17 Oct, 02:09 · ai-1

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

    17 Oct, 02:07 · engine:scheduler

11. New idea H-2df93a3d: Replicate: fell-going-well chase edge in IRE (replicate)

    17 Oct, 02:07 · 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.