A plain-English guide to our horse racing model, how we choose our picks, and our full published results record — updated daily.
Winsmore is a free, data-driven model for UK horse racing. Every day it ranks the runners in each race, publishes a pick per race, and records the result — win or lose. This page explains exactly how the model reaches its selections and shows our complete track record so you can judge it before you follow it.
Winsmore learns from the runners that have actually won. We build detailed profiles of past winners from a deep base of historical form — more than 90,000 UK races and over 800,000 individual runs since 2020, with Betfair pricing data going back to 2015 — and measure which factors most consistently separated winners from the field. The model never stops learning: it tracks more than 900 selections every month and recalibrates as each new result settles.
For every race on today's card, each runner is ranked against its rivals on every factor. Those rankings are combined using the model's learned importance for each factor to give a single model rating, and the highest-rated runner in the race becomes the model pick. It's a relative system: a horse is judged against the specific field it faces that day, not against an abstract benchmark.
The importance the model places on each factor is not fixed. It recalibrates every day as more results settle, so the model adapts over the season rather than assuming what worked in spring still works in autumn.
The rating blends the factors that most consistently separate winners from the field. The heaviest influence comes from a runner's recent form, its proven speed for the conditions, the strength of its trainer, and the weight of informed tips. Jockey record — including at the specific course — the draw, and a range of further race-specific factors then refine the picture. We keep the exact weighting proprietary, but the factors themselves are the ones any serious form student would recognise; our edge is in how they're measured, ranked and recalibrated daily.
The model produces one pick per race across the day's cards, and it prioritises conditions races for its strongest selections, where the form tends to be more reliable than in large-field handicaps. On top of the per-race picks, we publish three named selection tiers:
When no Dual Specialist qualifies on a given day, we publish our Top 3 Model Picks instead — simply the highest-rated back selections across the cards — so there is always a transparent selection on record.
We publish every selection and settle it at Betfair Starting Price. Nothing is hidden and losing runs stay on the record. The figures below are all-time to date and update daily on the site:
| Selection tier | Record | Strike rate | Profit/Loss | ROI |
|---|---|---|---|---|
| NAP (DS 6.0+ & top model pick) | 31/72 | 43% | +22.0 pts | +31% |
| Tooley's Tips (all DS 6.0+) | 56/179 | 31% | +31.4 pts | +18% |
| Backs (all model picks) | 695/2751 | 25% | +89.6 pts | +3% |
Strike rate is how often a selection wins. ROI is profit as a percentage of total stakes, to Betfair SP. Figures update as results settle; see the live results record for the current numbers.
The single biggest thing that beats us is by design. Because the model is built on form and history, it is at its weakest against lightly-raced and unexposed horses — first-time runners and rapid improvers whose ability isn't yet on the page. The model can only weigh evidence that exists, so a well-touted newcomer with no record can, and regularly does, beat a selection with a strong proven profile. It's the honest trade-off of a history-driven approach.
A few smaller patterns work against us too. Sundays are our weakest day, most likely because of the quality of racing on offer. Our Dual Specialist picks are seasonal: a typical month sees 30 or more, but summer months can be far thinner — driven by trainer and jockey combinations being absent over holidays and in extreme ground and heat — before volume recovers into autumn. And a winning long-run record still contains losing days: past performance is not a guarantee of future results.
Yes. Full cards, live prices, steamers, the daily model picks and the complete results record are all free on the site. There is no paywall on the selections.
No. Every selection is published and settled win or lose, and stays on the permanent record. We post our bad days as openly as our good ones — including runs like 0 from 11 — because a track record only means something if the losses are in it.
It ranks every runner in a race against its rivals on factors such as form, speed, trainer, jockey, tips and draw, multiplies each rank by that factor's weight, and picks the highest total. The weights are learned from past winners and update daily.
It's a tighter, higher-quality tier: a runner that qualifies on our Dual Specialist criteria with a DS rating of 6.0 or above. These are published as Tooley's Tips, and the best of them can become the day's NAP.
Selections are settled to Betfair Starting Price (BSP), so the recorded returns reflect prices that were actually available on the exchange rather than best-priced fractions.
All-time to date: Backs stand at +89.6 points (25% strike rate, +3% ROI), Tooley's Tips at +31.4 points (31%, +18% ROI), and the daily NAP at +22.0 points (43%, +31% ROI). These update daily on the site.