Threads · story · H24-one-off-jumping-error-well-beaten

H24: one off jumping error well beaten

Hypotheses: 27 (16 alive)
Depth: 5
Best rung: Live
Open questions: 20
Last activity: 17 Oct, 02:15

Where it stands: No: chase pull-ups with clean jumping that led last time score 0.68 on 43 (CI 0.14 to 1.97), the same as the whole clean pool (0.71 on 208). The broad led/prominent pull-up rival (H-a2870576) was killed, so run style does not concentrate the under-performance.

The tree

PROMISINGA/E 1.22 · 536H24: one off jumping error well beatenimportgreen

Still open

Why do previously poor jumpers by Jump Index outperform good ones inside this rule (1.879 vs 1.186 in the imported overlays)? open
Within the chase under-16/1 pool, is the lift in the mid-price band (around 5/1 to 14/1) rather than at the top of the market? open
Why do IRE hurdlers in this rule score 0.94 while GB score 1.16? Are big Irish maiden-hurdle fields making a 10L+ beating normal rather than misleading? open
If hurdles under 16/1 score only 1.08, where does the broad hurdle H24 lift (about 1.14) come from? Is it a 16/1+ longshot effect, such as improving novices in big fields, that has nothing to do with the jumping error? open
Why does the one-off error idea fail in IRE hurdles (0.944 on 222) but not in GB hurdles (1.16)? Do Irish comment-writers record hurdle jumping errors differently, so that last_jumping_error flags a different kind of run? open
Within the one-off chase-error close-run pool, is the lift carried by horses whose close run was a win (best_last3_beaten = 0), not a placed loss? testing
Is the lift among class risers (race_class < last_race_class) after a one-off chase error real, or two or three lucky winners on about 27 bets? open
Is the 16/1+ band in the GB no-drop chase pool (1.58 on 68, 3 wins) a real reversal of the earlier chase longshot failure, or noise from 3 winners? open
Why does the hurdle error pool pay only at 16/1+ (2.19 on 201, 8 wins) while the chase pool pays under 16/1? open
Why does the error edge reverse with margin: about 1.4 for runners beaten 10 to 20 lengths but 1.02 for those beaten more than 20?

The notebook

In order: what the lab saw, what it asked, what it proposed and why, what the tests said, and what an investigator concluded. Numbers are aggregates on the discovery data unless a step says otherwise.

Hypothesised15 Oct, 23:10
H24: one off jumping error well beaten PROMISING green
because A jumping error last time, clean in the two runs before, and well beaten: the mistake cost the run.
```
race_type IN ('Hurdle', 'Chase') AND remarked_last3 = 3 AND jumping_errors_last3 = 1 AND last_jumping_error = 1 AND (last_position IS NULL OR last_beaten_lengths > 10)
```
made by import
Tested15 Oct, 23:10
broad: A/E 1.17 on 1,570 bets, by period 1.12 · 1.23. info

Where it started

A jumping error last time, clean in the two runs before, and well beaten: the mistake cost the run.