Rival: the Snowden lift is a yard-wide effect; his non-novice jumps runners beat their prices too

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-610ab1e7 (Generalise: Snowden bumper runners beat their prices) passed discovery as a niche: A/E 1.11 on 35 bets (p 0.403), same-race control 0.94. O-c3563a6c
   • H-610ab1e7 (Generalise: Snowden bumper runners beat their prices) is unstable: 1 of 2 neighbouring rules lose the effect (base A/E 1.11; e.g. drop: trainer LIKE 'Jamie Snowden%' -> 0.99). O-ff3c493d
2. AskedIs the Snowden lift yard-wide (non-novices, handicaps, all codes) rather than specific to novices, so that 'novice' is a label on a general trainer effect? answered
3. MechanismA mid-sized yard that is in form, or improving its strike rate, may be underrated across the board because the market prices from its longer-run record. If so, the novice framing is incidental and the lift should show in non-novice hurdlers and chasers as well.
4. Why the market misses itBettors anchor on a trainer's long-run reputation and on each horse's visible form. A yard-level step up in this window would not yet be priced into runners whose form figures predate it.
5. Predictedpositive, A/E 1.04 to 1.10 on 200+ bets
6. Falsified ifA/E at or below 1.00 on 200+ bets, which would mean the lift really is novice-specific and the system's framing stands

Mechanism and rationale

A mid-sized yard that is in form, or improving its strike rate, may be underrated across the board because the market prices from its longer-run record. If so, the novice framing is incidental and the lift should show in non-novice hurdlers and chasers as well.

The rule

trainer LIKE 'Jamie Snowden%' AND is_novice = 0 AND is_maiden = 0 AND race_type IN ('Hurdle', 'Chase')

Reads: trainer, is_novice, is_maiden, race_type

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

fe6f44457738ed4c4ccb88ad85935e3ba9fbb8c4e7b9a31e56876ab3460df43f

Frozen

not frozen

The definition can still change.

Confounders

• Yard purple patch: a short in-form spell inside the window that would fade, not a lasting mispricing (check by season split)
• Handicap marks lagging improvement: if the lift sits only in handicaps, the mechanism is the rating system rather than yard reputation
• A few big-priced winners carrying A/E (check the 14/1-or-shorter refinement and winner count)
• Stable-jockey bookings: a lift tied to the stable jockey rather than the yard
• Edge only at BSP: the novice edge was +8.3% at BSP but -6.7% at SP, so an A/E lift may not be bettable at SP
• Overlap with the measured novice result: maidens are excluded so the two populations are disjoint

Control

{
  "description": "The other runners in the same races should sit near par (A/E about 0.97 to 1.00 after overround). If they also beat their prices, the lift belongs to the races Snowden picks, not to the yard. The second comparison is the yard's own novice and maiden jumps runners (the existing system population, about 1.11 on 488). This rule should be clearly positive but at or below that level. A result at or above the novice level would make the novice framing incidental. A result near 1.00 would confirm the lift is novice-specific.",
  "rule": "trainer LIKE 'Jamie Snowden%' AND (is_novice = 1 OR is_maiden = 1) AND race_type IN ('Hurdle', 'Chase')"
}

Discovery result

fail broad rule · 16 Oct, 15:18

Bets

187

34 winners

A/E

0.97

95%: 0.68 to 1.36

SP return

-28.3%

level stakes

Betfair SP return

-24.2%

178 bets with a BSP

Strike rate 18.2%. p = 0.58. 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 · 30s · 16 Oct, 15:17

  Output (schema-checked)

  {
    "rule": "trainer LIKE 'Jamie Snowden%' AND is_novice = 0 AND is_maiden = 0 AND race_type IN ('Hurdle', 'Chase')",
    "refinements": [
      {
        "label": "handicaps only (marks lagging yard improvement)",
        "rule": "trainer LIKE 'Jamie Snowden%' AND is_novice = 0 AND is_maiden = 0 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 1"
      },
      {
        "label": "14/1 or shorter (guard against a few big-priced winners carrying A/E)",
        "rule": "trainer LIKE 'Jamie Snowden%' AND is_novice = 0 AND is_maiden = 0 AND race_type IN ('Hurdle', 'Chase') AND sp_decimal <= 15"
      }
    ],
    "control": {
      "description": "The other runners in the same races should sit near par (A/E about 0.97 to 1.00 after overround). If they also beat their prices, the lift belongs to the races Snowden picks, not to the yard. The second comparison is the yard's own novice and maiden jumps runners (the existing system population, about 1.11 on 488). This rule should be clearly positive but at or below that level. A result at or above the novice level would make the novice framing incidental. A result near 1.00 would confirm the lift is novice-specific.",
      "rule": "trainer LIKE 'Jamie Snowden%' AND (is_novice = 1 OR is_maiden = 1) AND race_type IN ('Hurdle', 'Chase')"
    },
    "pass_criteria": {
      "min_bets": 200,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Yard purple patch: a short in-form spell inside the window that would fade, not a lasting mispricing (check by season split)",
      "Handicap marks lagging improvement: if the lift sits only in handicaps, the mechanism is the rating system rather than yard reputation",
      "A few big-priced winners carrying A/E (check the 14/1-or-shorter refinement and winner count)",
      "Stable-jockey bookings: a lift tied to the stable jockey rather than the yard",
      "Edge only at BSP: the novice edge was +8.3% at BSP but -6.7% at SP, so an A/E lift may not be bettable at SP",
      "Overlap with the measured novice result: maidens are excluded so the two populations are disjoint"
    ],
    "mechanism": "A mid-sized yard that is improving its strike rate is priced from its longer-run record. If so, the underrating applies to all its jumps runners, and the novice framing of the live Snowden system is incidental. Experienced (non-novice, non-maiden) hurdlers and chasers from the yard should then beat their prices too, perhaps by a little less than novices, whose prices rest on less visible form.",
    "market_miss": "Bettors anchor on a trainer's long-run reputation and each horse's visible form figures. A yard-level step up in this window is not yet priced into exposed runners whose form (and, in handicaps, official marks) predates it.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.04 to 1.10 on 200+ bets",
      "falsified_if": "A/E at or below 1.00 on 200+ bets. That would mean the lift is novice-specific and the system's framing stands. Also falsified if the lift exists only above 14/1 on a handful of winners."
    },
    "notes": "I kept the rule sketch and added one condition, is_maiden = 0. Maiden jumps runners are a novice-like group, so excluding them keeps this population separate from the measured novice and maiden result. Because the hypothesis predicts at least 200 bets, min_bets is raised to 200. I left out a non-handicap refinement: experienced runners are mostly in handicaps, so that slice would be thin and is already covered by the novice finding."
  }

Timeline

1. Job 1895 finished

   16 Oct, 15:18 · engine:compute-1

2. H-9672e9f3 Discovery → KILLED fails: 187 bets, needs 200; A/E 0.975, needs 1.05; p 0.5814, needs 0.05 or lower; periods [1.04, 0.84], each needs 1.0; A/E 0.975 against same-race control 1.012, needs 0.03 above (not a niche pass: A/E 0.975 below 1.05; p 0.5813593951815977 above 0.5; does not beat the same-race control 1.012; a period points the other way [1.04, 0.84])

   16 Oct, 15:18 · engine:evaluator

3. discovery result for H-9672e9f3 (broad rule): fail

   16 Oct, 15:18 · engine:evaluator

4. H-9672e9f3 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 15:18 · engine:orchestrator

5. compute-1 took job 1895 (evaluate, attempt 1)

   16 Oct, 15:18 · compute-1

6. Job 1894 finished

   16 Oct, 15:18 · engine:ai-1

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

   16 Oct, 15:18 · engine:scheduler

8. H-9672e9f3 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 15:18 · agent:designer@v1

9. ai-1 took job 1894 (design, attempt 1)

   16 Oct, 15:17 · ai-1

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

    16 Oct, 15:17 · engine:scheduler

11. New idea H-9672e9f3: Rival: the Snowden lift is a yard-wide effect; his non-novice jumps runners beat their prices too (rival-explanation)

    16 Oct, 15:17 · 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.