Wide on soft ground over jumps may be a chosen route, not lost ground

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
   • IQ140a-wide-trip-with-figures (IQ140a: wide trip with figures) was GREEN in the earlier research: A/E 1.27 on 1080 bets (periods 1.15 / 1.41); now PROMISING. O-943a9f88
2. AskedIn jump races on soft or heavy ground, is a wide trip a deliberate route to better ground, and does the edge shrink there? answered
3. MechanismOver jumps on soft or heavy ground, jockeys often steer wide on purpose for fresher ground. That horse gave up little or nothing, so the result is a fair reflection and there is nothing hidden to bet on.
4. Why the market misses itIf the rival is right there is no market miss here: the price is fair and these runners should be near 1.00, diluting the system. If the edge holds here too, the ground-lost reading is confirmed.
5. Predictednegative, A/E 0.95 to 1.05 on 150+ bets, clearly below the 1.27 overall
6. Falsified ifA/E 1.20 or higher on this subset, which would mean deliberate wide routes still undersell the horse
7. Rule sketchlast_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group IN ('soft', 'heavy')
8. question idQ-b6c8a25c

The question's answer so far: That subset is too small to matter: jump runners wide on soft or heavy ground last time number only 41 bets, at A/E 0.97. Whether the tag means deliberate routing there or not, it cannot explain the system's edge.

Mechanism and rationale

Over jumps on soft or heavy ground, jockeys often steer wide on purpose for fresher ground. That horse gave up little or nothing, so the result is a fair reflection and there is nothing hidden to bet on.

The rule

last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group IN ('soft', 'heavy')

Reads: last_raced_wide, last_position, last_beaten_lengths, last_slow_start, last_race_type, last_going_group

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

d5a3c3af7f6bbbc30c80aaeeeaa0bee3e03ac5b3c123fdc327c112c0903e760f

Frozen

not frozen

The definition can still change.

Confounders

• Winter jumps seasonality: soft and heavy ground falls mostly in November to March, so the subset is heavier in midwinter cards and bigger staying fields
• Thin window: last_raced_wide exists only from November 2025, so there is about one year in two halves, and soft-ground jumps is a fraction of that
• Soft-ground stayers trade at bigger prices, and the 16/1+ longshot lift common to beaten horses can inflate A/E (hence the 14/1 refinement)
• Today's race may be on different ground or a different code from the wide run, which mixes ground-change effects into the result
• Track mix: soft-ground jumps wide trips cluster at certain galloping tracks (e.g. Haydock, where wide routes are customary), so a track effect could pass for a ground effect
• Remark tagging: 'raced wide' in a jumps comment on soft ground may be written differently ('came wide for better ground'), so the tag may be uneven across going

Control

{
  "description": "Same-race other runners should be close to 1.00. The control that matters for the rival explanation is the same wide-trip trigger over jumps on ground that was NOT soft or heavy. Code is held the same and only the ground changes, so if the ground-routing explanation is right, that group should keep the system's edge (around 1.2 or higher) while the soft/heavy subset sits near 1.00. If both groups come out about the same, the ground-routing rival fails and the lost-ground reading stands.",
  "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group NOT IN ('soft', 'heavy')"
}

Discovery result

fail broad rule · 16 Oct, 11:16

Bets

41

7 winners

A/E

0.97

95%: 0.39 to 2.00

SP return

-30.9%

level stakes

Betfair SP return

-17.8%

41 bets with a BSP

Strike rate 17.1%. p = 0.58. Same-race control: the other runners in these races had A/E 1.01, so the selection did not beat its own races. Zone: discovery. Evaluator eval-1.0, data 2887c9abd664f46f

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 · 54s · 16 Oct, 11:15

  Output (schema-checked)

  {
    "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group IN ('soft', 'heavy')",
    "refinements": [
      {
        "label": "14/1 or shorter",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group IN ('soft', 'heavy') AND sp_decimal <= 15"
      },
      {
        "label": "heavy ground last time (where routing wide for fresher ground is most deliberate)",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group = 'heavy'"
      },
      {
        "label": "chase last time (fence lines make the outer route a common chosen line)",
        "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type = 'Chase' AND last_going_group IN ('soft', 'heavy')"
      }
    ],
    "control": {
      "description": "Same-race other runners should be close to 1.00. The control that matters for the rival explanation is the same wide-trip trigger over jumps on ground that was NOT soft or heavy. Code is held the same and only the ground changes, so if the ground-routing explanation is right, that group should keep the system's edge (around 1.2 or higher) while the soft/heavy subset sits near 1.00. If both groups come out about the same, the ground-routing rival fails and the lost-ground reading stands.",
      "rule": "last_raced_wide = 1 AND last_position > 1 AND last_beaten_lengths <= 5 AND last_slow_start = 0 AND last_race_type IN ('Hurdle', 'Chase') AND last_going_group NOT IN ('soft', 'heavy')"
    },
    "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": [
      "Winter jumps seasonality: soft and heavy ground falls mostly in November to March, so the subset is heavier in midwinter cards and bigger staying fields",
      "Thin window: last_raced_wide exists only from November 2025, so there is about one year in two halves, and soft-ground jumps is a fraction of that",
      "Soft-ground stayers trade at bigger prices, and the 16/1+ longshot lift common to beaten horses can inflate A/E (hence the 14/1 refinement)",
      "Today's race may be on different ground or a different code from the wide run, which mixes ground-change effects into the result",
      "Track mix: soft-ground jumps wide trips cluster at certain galloping tracks (e.g. Haydock, where wide routes are customary), so a track effect could pass for a ground effect",
      "Remark tagging: 'raced wide' in a jumps comment on soft ground may be written differently ('came wide for better ground'), so the tag may be uneven across going"
    ],
    "mechanism": "Over jumps on soft or heavy ground, jockeys often take a wide line on purpose to find fresher, less churned ground. Any extra distance is roughly paid back by the better footing. The horse then gave up little or nothing net, so its finishing position and beaten distance are a fair measure of its ability, and there is no hidden merit for the next race to reveal. If so, the wide-trip edge (A/E 1.27 overall) should shrink to about market level in this subset.",
    "market_miss": "Under this rival explanation there is no market miss in this subset. The form figures already read fairly, the price is fair, and these runners should sit near 1.00, which dilutes the published wide-trip system. If instead the edge holds at 1.20 or more here, the market is underrating the ground lost even on a chosen wide route, and the ground-lost reading of the system is confirmed.",
    "prediction": {
      "direction": "negative",
      "size": "A/E 0.95 to 1.05 on 150+ bets, clearly below the 1.27 of the overall wide-trip rule and below the jumps good-ground control",
      "falsified_if": "A/E 1.20 or higher on this subset, or no clear gap below the good-or-faster jumps control (within 0.05 of it). Either would mean deliberate wide routes on soft ground still undersell the horse."
    },
    "notes": "I checked the going values: last_going_group uses lower-case 'soft' and 'heavy', as other live rules do. I kept the draft rule unchanged because it matches the four conditions of the live system SYS-wide-trip-beaten-under-five, plus the jumps and soft-ground restriction the question needs. This is a rival-explanation (dilution) test. 'Negative' here means near or below the market and well under the system's 1.27, not necessarily a strong over-bet, so a 'pass' in the negative direction is unlikely by design. The decisive reading is the gap to the good-ground jumps control, not the absolute A/E. If the subset falls well short of 150 bets in the November 2025 window, treat the result as thin and descriptive. If it comes out at 1.20 or higher, that supports the ground-lost reading. Do not read it as a new trigger."
  }

Timeline

1. Job 1234 finished

   16 Oct, 11:16 · engine:compute-2

2. H-fc98577b Discovery → KILLED fails: 41 bets, needs 300; A/E 0.973, needs 0.95 or lower; p 0.57 (lower tail), needs 0.05; periods [], each needs 0.99 or lower

   16 Oct, 11:16 · engine:evaluator

3. discovery result for H-fc98577b (broad rule): fail

   16 Oct, 11:16 · engine:evaluator

4. H-fc98577b Designed → RESEARCHdiscovery measuring on the discovery zone

   16 Oct, 11:16 · engine:orchestrator

5. compute-2 took job 1234 (evaluate, attempt 1)

   16 Oct, 11:16 · compute-2

6. Job 1231 finished

   16 Oct, 11:15 · engine:ai-1

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

   16 Oct, 11:15 · engine:scheduler

8. H-fc98577b Proposed → RESEARCHdesigned pre-spec written and hashed

   16 Oct, 11:15 · agent:designer@v1

9. ai-1 took job 1231 (design, attempt 1)

   16 Oct, 11:15 · ai-1

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

    16 Oct, 11:14 · engine:scheduler

11. New idea H-fc98577b: Wide on soft ground over jumps may be a chosen route, not lost ground (rival-explanation)

    16 Oct, 11:14 · 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.