Test-mechanism: third favourites at 5/1 to 10/1 in big handicaps also beat their price

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-601a8566 (Rival: second favourites at 5/1+ in big handicaps also beat their pric) passed discovery as a niche: A/E 1.16 on 721 bets (p 0.087), same-race control 0.99. O-59b08a04
   • H-601a8566 (Rival: second favourites at 5/1+ in big handicaps also beat their pric): the country 'GB' carries the effect (A/E 1.4) while the rest score IRE 0.99. O-d291de7f
2. AskedDo third favourites at 5/1 to 10/1 in 14+ runner handicaps also beat their price, which would confirm a short-end cluster rather than something special about the top two? answered
3. MechanismIn a 14+ runner handicap the market's top few horses are the ones with real chances. Prices at 5/1 to 10/1 are calibrated mostly on fourth-to-sixth choices in smaller races, so the top choices of a big field get prices that understate how much the market concentrated on them. If this is the cause and not favouritism, the third choice should share it, perhaps more weakly.
4. Why the market misses itA/E benchmarks and bettors read a 7/1 shot as a 7/1 shot whatever its rank. A horse that is 7/1 third choice in a 16-runner handicap gets the same price as an unfancied 7/1 shot in an 8-runner race, though the market rated it much more highly relative to its field.
5. Predictedpositive, A/E 1.05 to 1.12 on 700+ bets
6. Falsified ifA/E 1.00 or below on 500+ bets, which would confine the effect to the top two and point back to a consensus mechanism

Mechanism and rationale

In a 14+ runner handicap the market's top few horses are the ones with real chances. Prices at 5/1 to 10/1 are calibrated mostly on fourth-to-sixth choices in smaller races, so the top choices of a big field get prices that understate how much the market concentrated on them. If this is the cause and not favouritism, the third choice should share it, perhaps more weakly.

The rule

is_handicap = 1 AND sp_rank = 3 AND sp_decimal >= 6 AND sp_decimal <= 11 AND field_size >= 14

Reads: is_handicap, sp_rank, sp_decimal, field_size

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

d1a378b9651a454d51acab77ad30eb34cdff0ae0c44ea98fa12d933e287f2662

Frozen

not frozen

The definition can still change.

Confounders

• GB/IRE mix: Ireland was at par for second favourites (0.99 on 436 bets) while GB scored 1.40, so a pooled figure may hide a GB-only effect. The GB-only refinement checks this.
• Joint-third favourites sharing a rank or tied prices: sp_rank = 3 may include co-favourite structures that do not match the 'clear third choice' story.
• Big festival handicaps (Cheltenham, Galway, Royal Ascot and similar) may dominate the sample and carry the result. Check the result by meeting and period.
• Overround differences between GB and Irish SP books: big-field books carry large overrounds, which compress top-end prices differently by jurisdiction.
• Price measure: H199's edge existed only at BSP (+16.6%) and not at SP (-9.6%), so A/E against SP may differ from returns that can actually be had.
• Overlap with the price band itself: the 5/1 to 10/1 range in big fields might be underbet whatever the rank. The rank-4+ control separates this.

Control

{
  "description": "Same-race other runners should sit at about A/E 0.97 to 1.00, as they did for the second-favourite test (0.99). The control that matters for the mechanism is horses at the same price band (5/1 to 10/1) in the same 14+ runner handicaps but ranked fourth or lower in the betting. The mechanism says rank within a big field, not the price itself, carries the edge, so this group should be near par (about 1.00) and clearly below the third favourites. If it shows the same lift, the effect is a price band in big handicaps rather than market concentration on the top choices.",
  "rule": "is_handicap = 1 AND sp_rank >= 4 AND sp_decimal >= 6 AND sp_decimal <= 11 AND field_size >= 14"
}

Discovery result

fail broad rule · 16 Oct, 15:49

Bets

1,246

123 winners

A/E

1.01

95%: 0.84 to 1.21

SP return

-25.3%

level stakes

Betfair SP return

-5.8%

1,200 bets with a BSP

Strike rate 9.9%. p = 0.46. Same-race control: the other runners in these races had A/E 1.01, 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 · 51s · 16 Oct, 15:47

  Output (schema-checked)

  {
    "rule": "is_handicap = 1 AND sp_rank = 3 AND sp_decimal >= 6 AND sp_decimal <= 11 AND field_size >= 14",
    "refinements": [
      {
        "label": "GB only",
        "rule": "is_handicap = 1 AND sp_rank = 3 AND sp_decimal >= 6 AND sp_decimal <= 11 AND field_size >= 14 AND region = 'GB'"
      },
      {
        "label": "Fields of 16+",
        "rule": "is_handicap = 1 AND sp_rank = 3 AND sp_decimal >= 6 AND sp_decimal <= 11 AND field_size >= 16"
      },
      {
        "label": "8/1 or shorter",
        "rule": "is_handicap = 1 AND sp_rank = 3 AND sp_decimal >= 6 AND sp_decimal <= 9 AND field_size >= 14"
      }
    ],
    "control": {
      "description": "Same-race other runners should sit at about A/E 0.97 to 1.00, as they did for the second-favourite test (0.99). The control that matters for the mechanism is horses at the same price band (5/1 to 10/1) in the same 14+ runner handicaps but ranked fourth or lower in the betting. The mechanism says rank within a big field, not the price itself, carries the edge, so this group should be near par (about 1.00) and clearly below the third favourites. If it shows the same lift, the effect is a price band in big handicaps rather than market concentration on the top choices.",
      "rule": "is_handicap = 1 AND sp_rank >= 4 AND sp_decimal >= 6 AND sp_decimal <= 11 AND field_size >= 14"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 150,
      "min_ae": 1.03,
      "max_p": 0.2
    },
    "confounders": [
      "GB/IRE mix: Ireland was at par for second favourites (0.99 on 436 bets) while GB scored 1.40, so a pooled figure may hide a GB-only effect. The GB-only refinement checks this.",
      "Joint-third favourites sharing a rank or tied prices: sp_rank = 3 may include co-favourite structures that do not match the 'clear third choice' story.",
      "Big festival handicaps (Cheltenham, Galway, Royal Ascot and similar) may dominate the sample and carry the result. Check the result by meeting and period.",
      "Overround differences between GB and Irish SP books: big-field books carry large overrounds, which compress top-end prices differently by jurisdiction.",
      "Price measure: H199's edge existed only at BSP (+16.6%) and not at SP (-9.6%), so A/E against SP may differ from returns that can actually be had.",
      "Overlap with the price band itself: the 5/1 to 10/1 range in big fields might be underbet whatever the rank. The rank-4+ control separates this."
    ],
    "mechanism": "In 14+ runner handicaps the market's money concentrates on the top few choices, which are the horses with realistic chances. Prices of 5/1 to 10/1 are usually read as they would be for fourth-to-sixth choices in smaller fields, so the top choices of a big field are priced as if the market concentrated on them less than it did. If the edge comes from this concentration effect at the short end of big fields and not from favourite status, the third choice at 5/1 to 10/1 should beat its price too, probably by less than the top two.",
    "market_miss": "Bettors and A/E-style benchmarks treat a 7/1 shot as a 7/1 shot whatever its rank. A 7/1 third choice in a 16-runner handicap is priced like an unfancied 7/1 shot in an 8-runner race, although the market rated it much more highly relative to its field. Big-field overround is also spread unevenly toward the top of the book. Neither effect is corrected, because rank-in-field is not part of how prices are read.",
    "prediction": {
      "direction": "positive",
      "size": "A/E 1.05 to 1.12 on 700+ bets, below the second-favourite figure of 1.16, with GB above Ireland",
      "falsified_if": "A/E of 1.00 or below on 500+ bets falsifies it, because that confines the effect to the top two and points to a consensus or favourite mechanism. It is also falsified if the rank-4+ control in the same price band scores as well as or better than the third favourites."
    },
    "notes": "The rule sketch is kept as written: all columns are pre-off and sp_decimal 6 to 11 covers 5/1 to 10/1. Refinements follow only from the reasoning and findings. GB-only follows from the GB-only second-favourite lift. 16+ runners follows from H235's stronger figure in bigger fields. 8/1 or shorter follows from the finding that the second-favourite sample sat at 5/1 to 7/1 and that edges thin out at longer prices. Pass min_bets is raised to 300 because the prediction expects 700+ bets and a modest A/E. The forward bar is set at 150 bets and A/E 1.03, and should be judged at BSP, given H199's finding that the edge exists only at exchange prices."
  }

Timeline

1. Job 1912 finished

   16 Oct, 15:49 · engine:compute-1

2. H-5c8f6618 Discovery → KILLED fails: A/E 1.011, needs 1.05; p 0.4634, needs 0.05 or lower; periods [0.98, 1.04], each needs 1.0; A/E 1.011 against same-race control 1.008, needs 0.03 above (not a niche pass: A/E 1.011 below 1.05; does not beat the same-race control 1.008)

   16 Oct, 15:49 · engine:evaluator

3. discovery result for H-5c8f6618 (broad rule): fail

   16 Oct, 15:49 · engine:evaluator

4. H-5c8f6618 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 15:49 · engine:orchestrator

5. compute-1 took job 1912 (evaluate, attempt 1)

   16 Oct, 15:49 · compute-1

6. Job 1910 finished

   16 Oct, 15:48 · engine:ai-1

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

   16 Oct, 15:48 · engine:scheduler

8. H-5c8f6618 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 15:48 · agent:designer@v1

9. ai-1 took job 1910 (design, attempt 1)

   16 Oct, 15:47 · ai-1

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

    16 Oct, 15:47 · engine:scheduler

11. New idea H-5c8f6618: Test-mechanism: third favourites at 5/1 to 10/1 in big handicaps also beat their price (test-mechanism)

    16 Oct, 15:47 · 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.