Big-rise quick-returning handicap winners stepping up 15%+ in trip are over-bet

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-7dbfdd08 (Rival: big-rise quick-returning handicap winners stepping up in class ) was predicted negative but came out the other way: A/E 1.16 on 138 bets (p 0.225 for the opposite direction). O-23b5bd4e
   • H-21f9e6e7 (Long-priced raised handicap winners turned out within 14 days are over) pointed the predicted way but failed the bar: A/E 0.57 on 255 bets (p 0.967), control 1.03. O-7c9eecf8
2. AskedDo big-rise quick returners at 16/1 or bigger share a visible obstacle (a class step-up, a trip change, a code switch) that explains why they fail? testing
3. MechanismA horse raised 6 lb or more must improve again to win. A big trip step-up asks for stamina it has not shown, on top of the extra weight, so many of these runs fail.
4. Why the market misses itPunters read a recent winner as progressive and assume it will stay the extra distance. The form book shows the win, not whether the horse can last the longer trip.
5. Predictednegative, A/E 0.75 to 0.90 on 100+ bets
6. Falsified ifA/E at or above 1.0 on 100+ bets. That would mean a trip step-up is no obstacle, and the longshot failure is a dismissal the market gets right for other reasons.
7. Rule sketchlast_position = 1 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change >= 6 AND distance_yards >= last_distance_yards * 1.15

Mechanism and rationale

A horse raised 6 lb or more must improve again to win. A big trip step-up asks for stamina it has not shown, on top of the extra weight, so many of these runs fail.

The rule

last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND race_type = last_race_type AND last_distance_yards IS NOT NULL AND distance_yards >= last_distance_yards * 1.15

Reads: last_position, days_since_last_run, is_handicap, last_is_handicap, rating_change, race_type, last_race_type, last_distance_yards, distance_yards

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

f29c562ec8fac71abdacfcf7ad940d46e8aaf8ecbb4b7879fe48bf46e4beea2f

Frozen

not frozen

The definition can still change.

Confounders

• Code switch: a Flat-to-hurdles move brings a large trip rise and a different rating scale, so the rule requires race_type = last_race_type
• Price band overlap: the parent group's failure sits at 16/1 or bigger, so a negative result may only reflect the price mix (check A/E within price bands against the control)
• Field size: longer handicaps, especially staying chases, have different field sizes
• Going change between the last run and today
• Class step-up that comes with the rise (an earlier finding showed it does not hurt this group)
• Thin window: rating_change only exists from 2025-11-01, so the sample may stay under 100

Control

{
  "description": "Same-race other runners should sit near A/E 1.0, so any shortfall belongs to the selected horses and not to the races. The key comparison is the same big-rise quick-returning handicap winners, same code, NOT stepping up 15% in trip. Earlier work puts that parent group at about 1.13 to 1.47. If the trip step-up is the real obstacle, the selection should sit clearly below this control, by 0.2 or more. If both are similar, trip is not the cause.",
  "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND race_type = last_race_type AND last_distance_yards IS NOT NULL AND distance_yards < last_distance_yards * 1.15"
}

Discovery result

fail broad rule · 16 Oct, 16:38

Bets

50

6 winners

A/E

0.67

95%: 0.25 to 1.47

SP return

-52.5%

level stakes

Betfair SP return

-44.0%

50 bets with a BSP

Strike rate 12.0%. p = 0.88. Same-race control: the other runners in these races had A/E 1.07, 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 · 42s · 16 Oct, 16:37

  Output (schema-checked)

  {
    "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND race_type = last_race_type AND last_distance_yards IS NOT NULL AND distance_yards >= last_distance_yards * 1.15",
    "refinements": [
      {
        "label": "8/1 or bigger: the longshot failure the question asks about",
        "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND race_type = last_race_type AND last_distance_yards IS NOT NULL AND distance_yards >= last_distance_yards * 1.15 AND sp_decimal >= 9"
      },
      {
        "label": "Marked step-up of 25% or more: a stronger stamina test should fail more often",
        "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND race_type = last_race_type AND last_distance_yards IS NOT NULL AND distance_yards >= last_distance_yards * 1.25"
      },
      {
        "label": "Stepping beyond the trip of its latest win: stamina truly unproven",
        "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND race_type = last_race_type AND last_distance_yards IS NOT NULL AND distance_yards >= last_distance_yards * 1.15 AND last_win_distance_yards IS NOT NULL AND distance_yards > last_win_distance_yards"
      }
    ],
    "control": {
      "description": "Same-race other runners should sit near A/E 1.0, so any shortfall belongs to the selected horses and not to the races. The key comparison is the same big-rise quick-returning handicap winners, same code, NOT stepping up 15% in trip. Earlier work puts that parent group at about 1.13 to 1.47. If the trip step-up is the real obstacle, the selection should sit clearly below this control, by 0.2 or more. If both are similar, trip is not the cause.",
      "rule": "last_position = 1 AND days_since_last_run IS NOT NULL AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change IS NOT NULL AND rating_change >= 6 AND race_type = last_race_type AND last_distance_yards IS NOT NULL AND distance_yards < last_distance_yards * 1.15"
    },
    "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": [
      "Code switch: a Flat-to-hurdles move brings a large trip rise and a different rating scale, so the rule requires race_type = last_race_type",
      "Price band overlap: the parent group's failure sits at 16/1 or bigger, so a negative result may only reflect the price mix (check A/E within price bands against the control)",
      "Field size: longer handicaps, especially staying chases, have different field sizes",
      "Going change between the last run and today",
      "Class step-up that comes with the rise (an earlier finding showed it does not hurt this group)",
      "Thin window: rating_change only exists from 2025-11-01, so the sample may stay under 100"
    ],
    "mechanism": "A handicap winner raised 6 lb or more and turned out within 14 days must improve again to win. A step-up of 15% or more in trip, in the same code, asks for stamina it has never shown, on top of the extra weight. Two demands that are not yet proven should make these runs fail more often than their prices imply. The quick-returning big-rise winners that do not step up in trip have beaten their prices (1.13 to 1.47).",
    "market_miss": "Punters read a quick-returning, freshly raised winner as progressive and assume it will stay a longer trip. The form figures show the win but not whether the horse can last the extra distance, so the stamina question is under-priced.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.75 to 0.90 on 100+ bets, at least 0.2 below the no-step-up control",
      "falsified_if": "A/E at or above 1.0 on 100+ bets, or no lower than the no-step-up control. That would mean a trip step-up is no obstacle, and the longshot failure is a price effect the market prices correctly."
    },
    "notes": "I added rating_change IS NOT NULL, days_since_last_run IS NOT NULL and last_distance_yards IS NOT NULL guards, which matches the family's earlier rules. I also added race_type = last_race_type: across codes a trip ratio and a rating change mean nothing (Flat and jumps official ratings are on different scales), and the hypothesis itself lists code switch as a confounder. rating_change is only available from 2025-11-01, so expect a thin sample that will likely fall short of 150 bets. In that case it should go on to falsification and forward tracking, not be killed. This is a negative-direction prediction, so the vault judges it as over-bet."
  }

Timeline

1. Job 2046 finished

   16 Oct, 16:38 · engine:compute-1

2. H-6ec43a26 Discovery → KILLED fails: 50 bets, needs 300; p 0.2145 (lower tail), needs 0.05; periods [], each needs 0.99 or lower

   16 Oct, 16:38 · engine:evaluator

3. discovery result for H-6ec43a26 (broad rule): fail

   16 Oct, 16:38 · engine:evaluator

4. H-6ec43a26 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:38 · engine:orchestrator

5. compute-1 took job 2046 (evaluate, attempt 1)

   16 Oct, 16:38 · compute-1

6. Job 2045 finished

   16 Oct, 16:38 · engine:ai-1

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

   16 Oct, 16:38 · engine:scheduler

8. H-6ec43a26 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:38 · agent:designer@v1

9. ai-1 took job 2045 (design, attempt 1)

   16 Oct, 16:37 · ai-1

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

    16 Oct, 16:37 · engine:scheduler

11. New idea H-6ec43a26: Big-rise quick-returning handicap winners stepping up 15%+ in trip are over-bet (rival-explanation)

    16 Oct, 16:37 · 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.