One horse. One bet. Every day. Sly Man’s read on the day’s card — the shortlist, the price, and the case for the pick.
Browse the logbookThe market told me something today before I even looked at the form — and on a card where half the shortlist has unavailable going data and blank form strings, that signal matters more than usual. The engine scanned all 35 races declared on 21 July 2026, applying a three-stage filter pipeline: first, dead or already-resulted races were cleared (none today); second, runners priced outside the 2.0–50.0 decimal odds band were removed — no odds-on, nothing at 50s or longer; third, races with fewer than six or more than fourteen declared runners were stripped out as too noisy or too chaotic. Feature scores were then built using weighted contributions across speed (1.5), market move (1.6), form (1.2), class change (1.0), going (0.8), trainer strike rate (0.8), and draw (0.5 — currently reserved for v2 and unavailable across the board).
The day started with 322 runners across 35 races. The odds-band filter trimmed that to 227, removing odds-on runners and any priced beyond 49/1. The field-size filter then cut to 158 runners in qualifying races — those with between six and fourteen declared starters. That is the universe the engine scored.
Five horses separated themselves at the top. High Tower landed the composite at 7.00/10, with Kaaranah just behind at 6.96/10 and Lady Felicia and Manoir De Mirande level on 6.88/10. Crystabel came in fifth at 6.61/10. What separated the top two was tight — four hundredths of a point — but the decisive gap came from the speed feature: High Tower's SR of 154 versus a field mean of 144.2 (z = +0.87) outpaced Kaaranah's raw SR of 72 once the heavier class environment at Salisbury was factored in alongside the missing-data penalties that clipped High Tower's floor score. The market-move feature — the highest-weighted in the engine at 1.6 — came in at +2.00 for both, so it did not separate them. Speed did.
The read I keep coming back to is the market move on High Tower. A strict majority of bookmakers shortened overnight, none drifted — that is the engine's steamer classification, scoring +2.00 on the 1.6-weighted market-move feature, and it is the single largest individual contribution to the 7.00/10 composite. At 5/2 (decimal 3.5), the engine estimates a true win probability of 35.6% against a market-implied 25.5% — that is an edge of +24.6%, which is real but not outrageous. It is the kind of number that says the market has not fully caught up, not that the market is obviously wrong. I find that more credible than a +285% edge on a 10/1 shot with one career run on record.
On ability, the speed rating tells a clear story. High Tower carries an SR of 154 into a field averaging 144.2 — a z-score of +0.87, contributing +0.56 weighted to the composite (speed weight 1.5, raw contribution +0.87). In a six-furlong novice stakes at Salisbury on good to firm, that raw pace advantage is the kind of edge that holds shape when the ground is quick and the race runs at a genuine clip. George Boughey posts a 15% trainer strike rate over the last 30 days from 9/62 — on par, not hot, contributing just +0.07 — so I am not leaning on the trainer angle. But I am not worried by it either.
The honest caveat in the case for High Tower is the volume of unavailable data: form string absent, class-change baseline absent, going record absent, draw absent. The composite is built on three live features — speed, trainer strike, and market move — rather than the full suite. The market-move score is doing significant heavy lifting here. My conviction rests on the combination of the steamer signal and the pace rating, not on a complete dossier. The engine returns MEDIUM conviction and so do I.
The unavailable data is the loudest risk. High Tower has no form string on file, no prior class comparison, no going record on good to firm or adjacent ground, and no draw assessment — four of seven features are either absent or defaulted to zero. That means the 7.00/10 composite is leaning hard on the speed rating and, above all, on the market-move signal. If the steam was bookmaker-led rather than genuine punter money — an SP management play rather than informed cash — the edge evaporates. The +24.6% estimated edge is meaningful but not commanding; Kaaranah at 6.96/10 is breathing down the neck of this selection and brings a fuller data picture including a live form string ('611651'), a hot trainer (John Butler at 29% from 10/34), and a speed z-score of +1.18. The counter-argument is that Kaaranah's 0-from-2 on good going gives me pause in conditions that matter. Going changes between now and the 18:30 at Wolverhampton could shift that calculus. At Salisbury, the good-to-firm ground is confirmed — but High Tower has no proven record on it.
High Tower · 5/2 · 1 unit · MEDIUM
The Fox’s Wire · Bet of the Day
The algorithm pulled five contenders from today’s card. One earned the ticket — the rest are what it beat. When each race settles, the finishes fill in so you can judge the shortlist, not just the selection.
Position only — payout-neutral. A process check, not a P/L. Click any horse to read its profile.
