Winners gifted an uncontested lead are over-bet at the front of the market next time

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-dd2d6d12 (Last-time winners that were slow in the race are over-bet at the front) pointed the predicted way but failed the bar: A/E 0.96 on 1399 bets (p 0.771), control 1.01. O-68e72211
2. AskedAre last-time winners whose win came from an uncontested, soft lead over-bet next time, when the speed figures alone are not? testing
3. MechanismA horse that is allowed an easy lead saves energy and wins by more than its ability merits. Next time rivals take it on, and the soft conditions rarely come again. The slow top-speed result was an indirect sign of this. An uncontested lead measures it directly.
4. Why the market misses itThe form book shows a clean win, often by a margin, and the rating rises. Punters back the winner without seeing that nobody pressed it. The pace context sits in the sectional data, not in the result line.
5. Predictednegative, A/E 0.85 to 0.93 on 400+ bets
6. Falsified ifA/E 0.96 or higher on 300+ bets, or no lower than the same-race control
7. Rule sketchlast_position = 1 AND l_uncontested = 1 AND sp_rank <= 3
8. question idQ-ac3df52b
9. operatortest-mechanism

Mechanism and rationale

A horse that is allowed an easy lead saves energy and wins by more than its ability merits. Next time rivals take it on, and the soft conditions rarely come again. The slow top-speed result was an indirect sign of this. An uncontested lead measures it directly.

The rule

last_position = 1 AND l_uncontested IS NOT NULL AND l_uncontested = 1 AND sp_rank <= 3

Reads: last_position, l_uncontested, sp_rank

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

9313edc997a8a8cd7c76c3d3b18d5904990f24ef84bf1343f542943787fb228a

Frozen

not frozen

The definition can still change.

Confounders

• Front-runner course bias: tracks that favour leaders give repeat easy leads, so the 'gift' may come back at the same course. Check whether the deficit holds when today's course differs from the last course.
• Small fields: an uncontested lead is common there, and the price may already allow for it. The last_field_size >= 8 refinement tests this.
• Handicap rating rise: the deficit could come from the raised mark rather than the pace setup. This overlaps the killed raised-winner result (0.84 at 9/1+). The non-handicap refinement isolates the pace effect.
• Generic low-figure or slow-winner effect: the killed slow-top-speed winner rule scored 0.96 on 1,399. The control of contested-lead winners has to show the uncontested group is worse, not merely below 1.
• Data window: l_uncontested only exists from GB 2023-04 and IRE 2024-01, so the sample is short and leans towards GB. Read GB and IRE separately.
• Coding of l_uncontested: the column is REAL, and if it is not a strict 0/1 flag, '= 1' may select only an extreme subset.

Control

Discovery result

fail broad rule · 16 Oct, 16:53

Bets

45

12 winners

A/E

0.93

95%: 0.48 to 1.62

SP return

-15.9%

level stakes

Betfair SP return

-1.7%

44 bets with a BSP

Strike rate 26.7%. p = 0.64. Same-race control: the other runners in these races had A/E 1.03, 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, 16:52

  Output (schema-checked)

  {
    "rule": "last_position = 1 AND l_uncontested IS NOT NULL AND l_uncontested = 1 AND sp_rank <= 3",
    "refinements": [
      {
        "label": "real field last time (8+ runners)",
        "rule": "last_position = 1 AND l_uncontested IS NOT NULL AND l_uncontested = 1 AND sp_rank <= 3 AND last_field_size >= 8"
      },
      {
        "label": "non-handicap today (removes the rating-rise confounder)",
        "rule": "last_position = 1 AND l_uncontested IS NOT NULL AND l_uncontested = 1 AND sp_rank <= 3 AND is_handicap = 0"
      },
      {
        "label": "handicap today, handicap win last time (rating-rise path)",
        "rule": "last_position = 1 AND l_uncontested IS NOT NULL AND l_uncontested = 1 AND sp_rank <= 3 AND is_handicap = 1 AND last_is_handicap = 1"
      }
    ],
    "control": {
      "description": "The usual same-race control: the other runners in the same races should come out near par (about 0.98 to 1.03). A second control separates the pace story from a general last-time-winner effect: last-time winners top three in the betting whose win was NOT from an uncontested lead. The mechanism needs the uncontested group to score clearly below this group (about 0.05 or more lower). If both groups sit at the same A/E, the deficit is a generic winner or short-price effect and not about pace.",
      "rule": "last_position = 1 AND l_uncontested IS NOT NULL AND l_uncontested = 0 AND sp_rank <= 3"
    },
    "pass_criteria": {
      "min_bets": 300,
      "min_ae": 1.05,
      "max_p": 0.05
    },
    "forward_bar": {
      "min_bets": 100,
      "min_ae": 1,
      "max_p": 0.2
    },
    "confounders": [
      "Front-runner course bias: tracks that favour leaders give repeat easy leads, so the 'gift' may come back at the same course. Check whether the deficit holds when today's course differs from the last course.",
      "Small fields: an uncontested lead is common there, and the price may already allow for it. The last_field_size >= 8 refinement tests this.",
      "Handicap rating rise: the deficit could come from the raised mark rather than the pace setup. This overlaps the killed raised-winner result (0.84 at 9/1+). The non-handicap refinement isolates the pace effect.",
      "Generic low-figure or slow-winner effect: the killed slow-top-speed winner rule scored 0.96 on 1,399. The control of contested-lead winners has to show the uncontested group is worse, not merely below 1.",
      "Data window: l_uncontested only exists from GB 2023-04 and IRE 2024-01, so the sample is short and leans towards GB. Read GB and IRE separately.",
      "Coding of l_uncontested: the column is REAL, and if it is not a strict 0/1 flag, '= 1' may select only an extreme subset."
    ],
    "mechanism": "A horse allowed to lead without pressure runs at an economical early pace. It saves energy and wins by more than its ability merits. Next time, rivals or a stronger pace take it on and the soft conditions rarely come again, so the winning form overstates its ability. The earlier slow-top-speed winner result (0.96) was an indirect proxy. RaceiQ's uncontested flag measures the pace gift directly.",
    "market_miss": "The result line shows a clean, often wide-margin win, and a handicap rating rises to match. Punters and the market make the horse one of the top three in the betting on that visible form. The fact that nobody pressed it sits in sectional and pace data that most bettors do not read.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.85 to 0.93 on 400+ bets, at least 0.05 below the contested-lead winner control",
      "falsified_if": "A/E 0.96 or higher on 300+ bets, or an A/E no lower than the same-race control, or no lower than the contested-lead winner control"
    },
    "notes": "The rule sketch is kept, with an IS NOT NULL guard added. It is evaluable only from GB 2023-04 and IRE 2024-01. min_bets is raised to 300 to match the falsification threshold. pass_criteria are read in the negative direction, as the lab does for this family. Refinements test the two confounders the reasoning names, field size and the handicap rating rise; no price cut is added because the mechanism is already limited to the front of the market."
  }

Timeline

1. Job 2065 finished

   16 Oct, 16:53 · engine:compute-1

2. H-07082d28 Discovery → KILLED fails: 45 bets, needs 300; p 0.4708 (lower tail), needs 0.05; periods [], each needs 0.99 or lower

   16 Oct, 16:53 · engine:evaluator

3. discovery result for H-07082d28 (broad rule): fail

   16 Oct, 16:53 · engine:evaluator

4. H-07082d28 Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 16:53 · engine:orchestrator

5. compute-1 took job 2065 (evaluate, attempt 1)

   16 Oct, 16:53 · compute-1

6. Job 2064 finished

   16 Oct, 16:52 · engine:ai-1

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

   16 Oct, 16:52 · engine:scheduler

8. H-07082d28 Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 16:52 · agent:designer@v1

9. ai-1 took job 2064 (design, attempt 1)

   16 Oct, 16:52 · ai-1

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

    16 Oct, 16:52 · engine:scheduler

11. New idea H-07082d28: Winners gifted an uncontested lead are over-bet at the front of the market next time (test-mechanism)

    16 Oct, 16:52 · 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.