Rival: any raised handicap winner turned out within 14 days beats its 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-97b1373c (Sharpen: easy handicap winners back within 14 days with no rise in the) pointed the predicted way but failed the bar: A/E 1.1 on 247 bets (p 0.208), control 0.95. O-d1dc9455
   • H-5c714267 (Test-mechanism: easy-winner quick-return lift lives in handicap-to-han) survived falsification: 5 of 5 attacks left it standing; it is frozen for its one vault query. O-e4c510dc
2. AskedIs the quick-return lift about the easy win at all once a handicap winner has been raised, or do all raised handicap winners turned out within 14 days beat their price? answered
3. MechanismA handicap winner raised and sent straight back is at the top of its form cycle whether or not it won easily. The market may punish the visible weight rise for all of them alike.
4. Why the market misses itPunters see 'up in the weights' and fade the horse. They underrate the form peak a quick turnout signals.
5. Predictedpositive, A/E 1.05 to 1.10 on 800+ bets if the rival is true; about 1.0 otherwise
6. Falsified ifA/E at or below 1.02 on 500+ bets, which would leave the easy-win flag as the active ingredient
7. Rule sketchlast_position = 1 AND last_won_easily = 0 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change > 0
8. question id

The question's answer so far: The lift needs the easy win. Raised handicap winners that did not win easily, back within 14 days, scored 1.034 on 1,050 (p 0.32), and that fell from 1.10 to 0.98 across the periods. Easy winners in the same frame scored 1.334.

Mechanism and rationale

A handicap winner raised and sent straight back is at the top of its form cycle whether or not it won easily. The market may punish the visible weight rise for all of them alike.

The rule

last_position = 1 AND last_won_easily = 0 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change > 0

Reads: last_position, last_won_easily, days_since_last_run, is_handicap, last_is_handicap, rating_change

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

a9cf604ca9f27238a3b62cfaac343a6461d185ff7d7dfebf41364961fe755d56

Frozen

not frozen

The definition can still change.

Confounders

• Trainer quality: big, in-form yards turn winners out quickly and are fairly priced or over-bet. Check the result is not carried by a few trainers.
• Class of the follow-up race: a raised winner stepping up in class differs from one staying at the same level. Compare race_class with last_race_class.
• Size of the rise: a small rise (1 to 3 lb) and a big reassessment (7 lb or more) could act in opposite directions.
• A quick turnout under a penalty versus after reassessment: within 14 days some rises are penalties and some are new marks, and rating_change may not show penalties the same way.
• Repeated ground or going, and same course and trip as the win.
• Field size and sp_rank: short-priced favourites dominate the bet count.

Control

{
  "description": "The lab measures the other runners in the same races. They should come out around 0.92 to 1.0, as they did for the easy-winner parent (control 0.92). The key comparison for this rival test is the matching easy-winner group: raised handicap-to-handicap easy winners back within 14 days. That group is given as the control rule. If the rival is true, the non-easy group should come close to the easy group, within about 0.1 A/E. If the non-easy group sits near 1.0 while the easy group sits well above it (about 1.2 or more, which the fixed-penalty finding implies), the easy-win flag is the active ingredient.",
  "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change > 0"
}

Discovery result

fail broad rule · 16 Oct, 16:22

Bets

1,050

206 winners

A/E

1.03

95%: 0.90 to 1.19

SP return

-18.6%

level stakes

Betfair SP return

-5.9%

1,038 bets with a BSP

Strike rate 19.6%. p = 0.32. Same-race control: the other runners in these races had A/E 0.99, 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 · 41s · 16 Oct, 16:21

  Output (schema-checked)

  {
    "rule": "last_position = 1 AND last_won_easily = 0 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change > 0",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "last_position = 1 AND last_won_easily = 0 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change > 0 AND sp_decimal <= 15"
      },
      {
        "label": "GB only",
        "rule": "last_position = 1 AND last_won_easily = 0 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change > 0 AND region = 'GB'"
      }
    ],
    "control": {
      "description": "The lab measures the other runners in the same races. They should come out around 0.92 to 1.0, as they did for the easy-winner parent (control 0.92). The key comparison for this rival test is the matching easy-winner group: raised handicap-to-handicap easy winners back within 14 days. That group is given as the control rule. If the rival is true, the non-easy group should come close to the easy group, within about 0.1 A/E. If the non-easy group sits near 1.0 while the easy group sits well above it (about 1.2 or more, which the fixed-penalty finding implies), the easy-win flag is the active ingredient.",
      "rule": "last_position = 1 AND last_won_easily = 1 AND days_since_last_run <= 14 AND is_handicap = 1 AND last_is_handicap = 1 AND rating_change > 0"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 150,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Trainer quality: big, in-form yards turn winners out quickly and are fairly priced or over-bet. Check the result is not carried by a few trainers.",
      "Class of the follow-up race: a raised winner stepping up in class differs from one staying at the same level. Compare race_class with last_race_class.",
      "Size of the rise: a small rise (1 to 3 lb) and a big reassessment (7 lb or more) could act in opposite directions.",
      "A quick turnout under a penalty versus after reassessment: within 14 days some rises are penalties and some are new marks, and rating_change may not show penalties the same way.",
      "Repeated ground or going, and same course and trip as the win.",
      "Field size and sp_rank: short-priced favourites dominate the bet count."
    ],
    "mechanism": "A handicap winner sent straight back within a fortnight with a higher official rating is at the top of its form cycle, whether or not it won easily. Its current ability may still be ahead of the new mark. If the quick-return lift is about form peak rather than how easily it won, non-easy raised winners should beat their price too.",
    "market_miss": "Punters see 'up in the weights' on a recent winner and fade it. They do not credit the form peak that a quick turnout signals. Non-easy winners carry no 'won with plenty in hand' comment to offset that fade, so if the fade is wrong, they should be the most underrated.",
    "prediction": {
      "direction": "positive",
      "size": "A/E about 1.05 to 1.10 on 800+ bets if the rival explanation is true; about 1.0 if not",
      "falsified_if": "A/E at or below 1.02 on 500 or more bets, or an A/E more than 0.15 below the raised easy-winner control group. Either leaves the easy-win flag as the active ingredient."
    },
    "notes": "rating_change and last_won_easily are only available from 2025-11-01 (GB and IRE), so the sample window is short. The rule is kept exactly as sketched. The earlier broad rival (all non-easy last-time winners within 14 days) scored 1.00 on 2,015, so this tests only the narrower raised handicap-to-handicap claim. The refinements come from family findings: the long-price end did nothing (0 wins from 24 at 16/1 or bigger) and the lift was GB-only."
  }

Timeline

1. Job 1981 finished

   16 Oct, 16:22 · engine:compute-1

2. H-a4582701 Discovery → KILLED fails: A/E 1.034, needs 1.05; p 0.3228, needs 0.05 or lower; periods [1.1, 0.98], each needs 1.0 (not a niche pass: A/E 1.034 below 1.05)

   16 Oct, 16:22 · engine:evaluator

3. discovery result for H-a4582701 (broad rule): fail

   16 Oct, 16:22 · engine:evaluator

4. H-a4582701 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:22 · engine:orchestrator

5. compute-1 took job 1981 (evaluate, attempt 1)

   16 Oct, 16:22 · compute-1

6. Job 1979 finished

   16 Oct, 16:22 · engine:ai-1

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

   16 Oct, 16:22 · engine:scheduler

8. H-a4582701 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:22 · agent:designer@v1

9. ai-1 took job 1979 (design, attempt 1)

   16 Oct, 16:21 · ai-1

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

    16 Oct, 16:20 · engine:scheduler

11. New idea H-a4582701: Rival: any raised handicap winner turned out within 14 days beats its price (rival-explanation)

    16 Oct, 16:20 · 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.