Rival: Snowden novice lift is a short-price artefact; test the 9/2+ runners on their own

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
   • H17-jamie-snowden-novices (H17: jamie snowden novices) was GREEN in the earlier research: A/E 1.11 on 488 bets (periods 1.11 / 1.07 / 1.20); now PROMISING. O-71eb5f8a
2. AskedIs the Snowden novice non-handicap lift spread across the price range, or carried by one short-price band the way the marathon slices were? testing
3. MechanismIf the yard really produces novices who improve more than their form shows, the market underrates them at every price, and most of all when they are not strongly fancied. If the lift is only at short prices, it is the general short-price cluster seen elsewhere, not a trainer trait.
4. Why the market misses itRunners from a mid-sized yard at 9/2 or bigger in novice races are priced off thin, unflattering early form. The improvement from schooling and from moving up in trip does not show in the form book.
5. Predictedpositive, A/E 1.08 to 1.20 on 150+ bets
6. Falsified ifA/E at or below 1.00 on 150+ bets at SP 5.5 or bigger, which would mean the system rides on the short-price band alone
7. Rule sketchtrainer LIKE 'Jamie Snowden%' AND is_novice = 1 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 0 AND sp_decimal >= 5.5
8. question idQ-14b21131

Mechanism and rationale

If the yard really produces novices who improve more than their form shows, the market underrates them at every price, and most of all when they are not strongly fancied. If the lift is only at short prices, it is the general short-price cluster seen elsewhere, not a trainer trait.

The rule

trainer LIKE 'Jamie Snowden%' AND is_novice = 1 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 0 AND sp_decimal >= 5.5

Reads: trainer, is_novice, race_type, is_handicap, sp_decimal

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

59fd2e4d10f417a0a0ffba2501eca061ecb9cf904c36889373fe29f624343aa8

Frozen

not frozen

The definition can still change.

Confounders

• The general short-price cluster seen in other families (up-to-7/1 band carrying slices). This rule excludes sub-5.5 runners, but the 5.5 to 8.0 band could still carry it, so the result should be broken down by price band.
• Few winners at long prices: one or two 20/1+ winners can make A/E, so check the result with the biggest winner removed.
• Festival or graded (Class 1) novice races, where prices are tighter and fields stronger. These are handled by a refinement.
• Overlap with the existing LIVE Snowden non-handicap novice system (282 bets). This is a subset of it, not independent support.
• Edge visible only at BSP (+8.3% at BSP, -6.7% at SP), so an SP-based A/E may differ from what can actually be taken.
• Yard form cycles (hot or cold spells) bunching winners in a few months.

Control

{
  "description": "The lab's same-race comparison: other runners in these non-handicap novice jumps races. Expect them near par or a little below (about 0.92 to 1.00), as mid- and long-priced runners usually are after the longshot bias. If those runners also beat their prices, the lift belongs to the race type and not to the yard. A second group, given as the control rule, is every other trainer's non-handicap novice jumps runner at 5.5 or bigger. This isolates the trainer effect at the same price band, and it should sit at about 0.90 to 0.98. For a descriptive comparison, the lab should also report Snowden's own sub-5.5 non-handicap novice runners. The rival explanation predicts those carry the 1.17 while this rule sits at or below 1.00.",
  "rule": "trainer NOT LIKE 'Jamie Snowden%' AND is_novice = 1 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 0 AND sp_decimal >= 5.5"
}

Discovery result

fail broad rule · 16 Oct, 15:16

Bets

27

3 winners

A/E

1.43

95%: 0.29 to 4.19

SP return

+20.4%

level stakes

Betfair SP return

–

27 bets with a BSP

Strike rate 11.1%. p = 0.35. Same-race control: the other runners in these races had A/E 0.96, 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 · 32s · 16 Oct, 15:15

  Output (schema-checked)

  {
    "rule": "trainer LIKE 'Jamie Snowden%' AND is_novice = 1 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 0 AND sp_decimal >= 5.5",
    "refinements": [
      {
        "label": "9/2 to 14/1 band only",
        "rule": "trainer LIKE 'Jamie Snowden%' AND is_novice = 1 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 0 AND sp_decimal >= 5.5 AND sp_decimal <= 15.0"
      },
      {
        "label": "exclude Class 1 (graded/festival) novice races",
        "rule": "trainer LIKE 'Jamie Snowden%' AND is_novice = 1 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 0 AND sp_decimal >= 5.5 AND (race_class IS NULL OR race_class > 1)"
      }
    ],
    "control": {
      "description": "The lab's same-race comparison: other runners in these non-handicap novice jumps races. Expect them near par or a little below (about 0.92 to 1.00), as mid- and long-priced runners usually are after the longshot bias. If those runners also beat their prices, the lift belongs to the race type and not to the yard. A second group, given as the control rule, is every other trainer's non-handicap novice jumps runner at 5.5 or bigger. This isolates the trainer effect at the same price band, and it should sit at about 0.90 to 0.98. For a descriptive comparison, the lab should also report Snowden's own sub-5.5 non-handicap novice runners. The rival explanation predicts those carry the 1.17 while this rule sits at or below 1.00.",
      "rule": "trainer NOT LIKE 'Jamie Snowden%' AND is_novice = 1 AND race_type IN ('Hurdle', 'Chase') AND is_handicap = 0 AND sp_decimal >= 5.5"
    },
    "pass_criteria": {
      "min_bets": 150,
      "min_ae": 1.08,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1.03,
      "max_p": 0.2
    },
    "confounders": [
      "The general short-price cluster seen in other families (up-to-7/1 band carrying slices). This rule excludes sub-5.5 runners, but the 5.5 to 8.0 band could still carry it, so the result should be broken down by price band.",
      "Few winners at long prices: one or two 20/1+ winners can make A/E, so check the result with the biggest winner removed.",
      "Festival or graded (Class 1) novice races, where prices are tighter and fields stronger. These are handled by a refinement.",
      "Overlap with the existing LIVE Snowden non-handicap novice system (282 bets). This is a subset of it, not independent support.",
      "Edge visible only at BSP (+8.3% at BSP, -6.7% at SP), so an SP-based A/E may differ from what can actually be taken.",
      "Yard form cycles (hot or cold spells) bunching winners in a few months."
    ],
    "mechanism": "If Snowden's yard produces novices who improve more than their form shows (from schooling and from stepping up in trip), the market underrates them at every price, and most of all when they are not strongly fancied. A real trainer trait should show at 9/2 and bigger, not only near the head of the market. If the lift exists only below 9/2, it is the general short-price cluster seen across families and not a trait of the yard.",
    "market_miss": "Runners at 9/2 or bigger from a mid-sized yard in non-handicap novice races are priced off thin, unflattering early form, with no official rating to anchor them. The yard's improvement from schooling and moving up in trip does not show in the form book, and attention and money go to the fancied runners from bigger yards.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.08 to 1.20 on roughly 150 to 200 bets",
      "falsified_if": "A/E at or below 1.00 on 150+ bets at SP 5.5 or bigger. That would mean the Snowden non-handicap novice system rides on the short-price band alone. The same applies if the lift comes from the 5.5 to 8.0 band alone while the rest sits below par."
    },
    "notes": "The draft rule is sound and kept unchanged. All its columns are pre-off and available from 2020-07. The sample is expected to be thin (the parent set is 282 non-handicap bets, and probably about half of those are at 5.5 or bigger). Per lab policy, a thin result that points the right way goes on to falsification rather than being killed. The minimum A/E is raised to 1.08 to match the predicted lower bound. The lab should report the result broken down by band (5.5 to 8, 8 to 15, 15+) and with the top winner removed. Evaluate at both BSP and SP given the earlier finding."
  }

Timeline

1. Job 1888 finished

   16 Oct, 15:16 · engine:compute-1

2. H-0c98b030 Discovery → KILLED fails: 27 bets, needs 150; p 0.3487, needs 0.05 or lower; periods [], each needs 1.0 (not a niche pass: only 27 bets)

   16 Oct, 15:16 · engine:evaluator

3. discovery result for H-0c98b030 (broad rule): fail

   16 Oct, 15:16 · engine:evaluator

4. H-0c98b030 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 15:16 · engine:orchestrator

5. compute-1 took job 1888 (evaluate, attempt 1)

   16 Oct, 15:16 · compute-1

6. Job 1885 finished

   16 Oct, 15:15 · engine:ai-1

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

   16 Oct, 15:15 · engine:scheduler

8. H-0c98b030 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 15:15 · agent:designer@v1

9. ai-1 took job 1885 (design, attempt 1)

   16 Oct, 15:15 · ai-1

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

    16 Oct, 15:15 · engine:scheduler

11. New idea H-0c98b030: Rival: Snowden novice lift is a short-price artefact; test the 9/2+ runners on their own (rival-explanation)

    16 Oct, 15:15 · 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.