Sly Man’s Bet of the Day

Method

Short-priced steamers with genuine speed behind them are the ones I trust — and today that combination narrows the field fast. I ran 364 declared runners across 34 races through a four-stage pipeline: odds band 2.0–50.0 decimal (no odds-on, no 50/1 longshots), field size 6–14 runners to cut out the noise at both ends, then a Lay of the Day collision check that dropped a further 2 runners. What survived went into a weighted composite built around seven features — market_move carrying the heaviest load at 1.6, speed at 1.5, form at 1.2, class_change and going both at 0.8, trainer_strike at 0.8, and draw reserved for v2. Today's engine conviction is MEDIUM.

Today's universe

The day opened with 364 runners across 34 races. The odds filter cut that to 320 — removing odds-on shots and anything beyond 50.0 decimal. Field size brought it down to 246, keeping only races with 6–14 declared runners. The Lay of the Day collision check removed a further 2, leaving 244 runners entering the composite scoring stage.

The shortlist

Five horses pushed into meaningful territory. Boilermaker topped the board at 7.42/10, a nose ahead of Bidsta at 7.40/10 — the closest finish on the shortlist. Rocky's Howya came in third at 7.09/10, followed by Empirical at 7.05/10 and Yabreen at 6.96/10. Wezzeer rounded out the six at 6.95/10. What separated the top two was market_move — both maxed the feature at +2.00 — so it came down to speed: Boilermaker posted a z-score of +1.82 against a field mean of 69.3, versus Bidsta's +1.28. That gap, combined with a readable form string for Boilermaker where Bidsta had none on file, tipped the balance.

Why this horse

The read I keep coming back to is the market_move score — Boilermaker hit the ceiling at +2.00, meaning a strict majority of bookmakers shortened on the morning line with none going the other way. That is the cleanest possible signal this engine recognises, and it carries the highest feature weight at 1.6. But unlike some steamers that ride purely on money, the speed data here is substantive: an SR of 88 against a field mean of 69.3 puts Boilermaker at a z-score of +1.82 — that is the widest speed margin on today's full shortlist.

The form string '-42321' is encouraging too. The decay-weighted score of +1.79 reflects a horse trending in the right direction — a recent win, placed runs on either side, and nothing alarming in the sequence. Harry Charlton's yard is running hot at 21% over the last 30 days (6 from 29), which adds a quiet layer of confidence. And the class drop — from rank 5 last time out to rank 6 today — is a small but real edge in a Class 5 context.

The one honest caveat on the market side: the engine's estimated win probability is 37.6% against a market-implied 42.5%, which shows a negative edge of -21.0%. That matters. The model thinks the market has slightly overpriced this horse. I am backing the steamer signal and the speed gap anyway, but I am not pretending the model edge is on my side here — the conviction is MEDIUM, not high, and that pricing gap is part of the reason.

What worries me

The negative market edge of -21.0% is the number I cannot ignore — the market implies a higher chance of winning than the model's own estimate, which means the price at 11/10 (decimal 2.1) is not the value play some steamers are. Going is listed as Good and Boilermaker is 1-from-4 with three places on good or adjacent ground — solid rather than dominant. The draw-bias data layer is unavailable for v2, so on an AW track at Wolverhampton over 1m5f219y I am flying partially blind on positional bias. Bidsta at 7.40/10 is only 0.02 behind and is also a confirmed steamer — if that race sets up differently to how I expect, the margin between selection and runner-up looks very thin.

The bet

Boilermaker · 11/10 · 1 unit · MEDIUM