H19: momentum interrupted beaten narrowly

Mechanism and rationale

Met trouble while making its move last time and was still beaten 2 lengths or less.

The rule

last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2

Reads: last_momentum_interrupted, last_position, last_beaten_lengths

Pre-registration

Spec hash (sha256 of the canonical pre-spec)

not designed yet

Frozen

not frozen

The definition can still change.

Broad

last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2

Kind

remarks

Origin

research/experiments/overnight/results_v2.jsonl

Refinements

[
  {
    "name": "Flat",
    "rule": "(last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2) AND (race_type = 'Flat')"
  },
  {
    "name": "handicap",
    "rule": "(last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2) AND (is_handicap = 1)"
  },
  {
    "name": "not a flattered closer",
    "rule": "(last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2) AND (coalesce(last_race_tilt, 0) < 0.2)"
  }
]

Tier

A

Result (from the archive)

info not a flattered closer · 15 Oct, 23:10

Bets

218

A/E

1.28

winners vs the prices

SP return

+15.8%

level stakes

Betfair SP return

+29.8%

216 bets with a BSP

Strike rate 22.5%.

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

• investigator v1ok · 47s · 16 Oct, 14:44

  Output (schema-checked)

  {
    "reflection": "The thread rests on one idea: trouble costs a horse only when it comes while the horse is making its move, and a horse checked mid-run that still finished within 2 lengths ran better than the result shows. The earlier research rated it GREEN at A/E 1.18 on 242 bets. It was above par in both periods (1.07 on 96, then 1.26 on 146), so the lift did not fade, and it is now PROMISING. Discovery gave A/E 1.28 on 218, with the strongest reading on the Flat. The BSP ROI of 19.1% is far above the SP ROI of 6.4%, so much of the value goes to the bookmakers' margin at SP. The live system adds a filter that drops races that flattered closers (last_race_tilt < 0.2), but that filter has not been measured on its own. The chain has not yet shown which part does the work: the trouble, or simply being beaten narrowly. It also has not shown whether the effect reaches beyond the Flat. Those are the next tests. Slicing the 242 bets further would only chase noise.",
    "answers": [],
    "questions": [
      {
        "question": "Is the lift caused by the interrupted run, or would any horse beaten 2 lengths or less (but not first) beat its price just as well?",
        "rationale": "If narrowly beaten horses that met no trouble also show A/E near 1.18, the trouble flag adds nothing and the system is really a 'narrow loser' rule, which the market should already price. Columns: last_momentum_interrupted, last_position, last_beaten_lengths, sp_decimal.",
        "observation_ids": [
          "O-a6c61d14"
        ],
        "priority": 0.7
      },
      {
        "question": "Does the momentum-interrupted, narrowly-beaten lift hold in hurdles and chases, or only on the Flat where discovery was strongest?",
        "rationale": "On the Flat, races are short and tactical, so a check mid-run is decisive. Over jumps, in longer races, horses have more time to recover. Columns: race_type, last_race_type, last_momentum_interrupted, last_beaten_lengths.",
        "observation_ids": [
          "O-a6c61d14"
        ],
        "priority": 0.5
      },
      {
        "question": "Is the edge biggest when the market ignores the excuse, i.e. when today's SP is no shorter than the last SP?",
        "rationale": "If the market reads the trouble comment, it will shorten these horses, and the value should sit with those that did not shorten. Columns: sp_decimal, last_sp_decimal, sp_rank.",
        "observation_ids": [
          "O-a6c61d14"
        ],
        "priority": 0.4
      }
    ],
    "hypotheses": [
      {
        "title": "Rival: narrowly beaten runners with no trouble last time beat their price as well",
        "family": "story",
        "operator": "rival-explanation",
        "question_id": null,
        "new_question": 0,
        "parent_id": "H19-momentum-interrupted-beaten-narrowly",
        "reasoning": {
          "observation_ids": [
            "O-a6c61d14"
          ],
          "question": "Does narrow defeat alone produce the H19 lift, without any trouble in running?",
          "mechanism": "A horse beaten 2 lengths or less may simply be in form and close to winning, whatever happened in the run. If so, the lift in H19 comes from the narrow margin, not from the interruption.",
          "market_miss": "If this rival holds, the market underrates narrow losers in general. That is unlikely, because a close second or third is the most visible line in the form book. So a par result here would support H19's trouble mechanism.",
          "prediction": {
            "direction": "positive",
            "size": "A/E 1.10 or more on 3000+ bets if the rival is true; about 1.00 if H19's mechanism is right",
            "falsified_if": "A/E 1.03 or below on 3000+ bets, which rules out the rival and leaves the trouble flag as the source of the edge"
          },
          "rule_sketch": "last_momentum_interrupted = 0 AND last_position > 1 AND last_beaten_lengths <= 2"
        },
        "confounders": [
          "field size of last race",
          "last_sp_decimal (short-priced narrow losers)",
          "race_type mix"
        ],
        "expected_survival": 0.25
      },
      {
        "title": "Replicate: momentum interrupted, beaten 2 lengths or less, in hurdles and chases",
        "family": "story",
        "operator": "replicate",
        "question_id": null,
        "new_question": 1,
        "parent_id": "H19-momentum-interrupted-beaten-narrowly",
        "reasoning": {
          "observation_ids": [
            "O-a6c61d14"
          ],
          "question": "Does the H19 lift also hold over jumps?",
          "mechanism": "A horse checked while making its move loses momentum it cannot fully win back. Over jumps a check is less decisive than on the Flat, but a horse that still finished within 2 lengths should still have more ability than the result shows.",
          "market_miss": "The result shows a beaten horse. Jumps punters weigh jumping errors more than checks in running, so a check may be under-read.",
          "prediction": {
            "direction": "positive",
            "size": "A/E 1.05 to 1.15 on 150+ bets",
            "falsified_if": "A/E below 1.0 on 150+ bets, or a lift in only one period"
          },
          "rule_sketch": "last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND race_type IN ('Hurdle','Chase')"
        },
        "confounders": [
          "small sample",
          "jumping errors confused with interruption",
          "going"
        ],
        "expected_survival": 0.35
      },
      {
        "title": "Test-mechanism: the edge sits with horses the market did not shorten after the excuse",
        "family": "story",
        "operator": "test-mechanism",
        "question_id": null,
        "new_question": 2,
        "parent_id": "H19-momentum-interrupted-beaten-narrowly",
        "reasoning": {
          "observation_ids": [
            "O-a6c61d14"
          ],
          "question": "Is the value concentrated where the price shows no allowance for the trouble?",
          "mechanism": "If H19 works because the market misses the excuse, value should be largest where the price did not shorten from the last run. Where the market did shorten, it has already priced the excuse.",
          "market_miss": "Bookmakers price from the finishing position and the margin. The in-running interruption is only in comments, and many punters skip them.",
          "prediction": {
            "direction": "positive",
            "size": "A/E 1.20 or more on 100+ bets, above the shortened group",
            "falsified_if": "A/E at or below that of the H19 runners whose price shortened, or below 1.05"
          },
          "rule_sketch": "last_momentum_interrupted = 1 AND last_position > 1 AND last_beaten_lengths <= 2 AND sp_decimal >= last_sp_decimal"
        },
        "confounders": [
          "class rise today",
          "trip or going change driving the drift",
          "longshot noise"
        ],
        "expected_survival": 0.35
      }
    ],
    "findings": [
      {
        "statement": "The rule for horses whose run was interrupted last time and who were beaten 2 lengths or less beat its price in both earlier periods (A/E 1.07 on 96, then 1.26 on 146), so the lift did not fade.",
        "kind": "holds",
        "observation_ids": [
          "O-a6c61d14"
        ]
      },
      {
        "statement": "The rule's BSP ROI (19.1%) is about three times its SP ROI (6.4%), so the edge is only worth taking at exchange prices.",
        "kind": "data",
        "observation_ids": [
          "O-a6c61d14"
        ]
      }
    ]
  }

Timeline

1. Job 56 finished

   15 Oct, 23:29 · engine:compute-2

2. compute-2 took job 56 (mutate, attempt 1)

   15 Oct, 23:29 · compute-2

3. Queued mutate on the compute lane: AI unavailable: deterministic mutations of a live candidate

   15 Oct, 23:29 · engine:scheduler

4. hypothesis note H19-momentum-interrupted-beaten-narrowly

   15 Oct, 23:10 · importer

5. H19-momentum-interrupted-beaten-narrowly Discovery → PROMISING import: tier A: strong candidate; best [Flat] GREEN A/E 1.28 on 218 bets (broad GREEN)

   15 Oct, 23:10 · importer

6. import result for H19-momentum-interrupted-beaten-narrowly (not a flattered closer): info

   15 Oct, 23:10 · importer

7. import result for H19-momentum-interrupted-beaten-narrowly (handicap): info

   15 Oct, 23:10 · importer

8. import result for H19-momentum-interrupted-beaten-narrowly (Flat): info

   15 Oct, 23:10 · importer

9. import result for H19-momentum-interrupted-beaten-narrowly (broad): info

   15 Oct, 23:10 · importer

10. New idea H19-momentum-interrupted-beaten-narrowly: H19: momentum interrupted beaten narrowly (import)

    15 Oct, 23:10 · importer

Every change is an event with an actor: an agent and its version, an engine, a human, or the importer. Current stage: Promising.