A model, not a man with a tip
Every Algohorse selection comes from a machine-learning model trained on decades of racing — graded the same way, every day. Here's the life of a tip, in three steps.
The model rates every runner
38 years of results, form, ratings, going, draw and connections — 3.8M runner records across 400,000+ races in UK and Irish racing — feed a gradient-boosted model that puts a calibrated win and place probability on every horse, every morning. No hunches, no narratives.
Registered before the off
Selections are tiered (NAP, Next Best, Speculative) and logged at the price available — before the race, so nothing is hindsight. What you see each morning is exactly what goes on the permanent record.
Settled in public
After racing, every tip is graded automatically: placed or not, and whether the advised price beat the SP (CLV). Good day or bad, the record updates the same way — that's the whole point.
Why this is different from a tipster
A tipster gives you an opinion. A model gives you a process — the same inputs, the same maths, every single day, with no room for ego, recency bias or a good story about a well-bred maiden. When the model is wrong, you can see why, and the next day it still does exactly what it did before.
Under the hood, two things keep the numbers honest. First, the raw model scores are calibrated against outcomes, so the ordering is stable rather than drifting with the model's mood. We used to publish that probability on every pick and we stopped: on our own live tips it carried no usable information about which of them would place, so quoting it was giving members a number that felt precise and wasn't. Second, the model is walk-forward validated: trained on the past, judged on future seasons it never saw, with strict controls so no post-race information ever leaks into a feature. If a version of the model can't beat that test, it doesn't ship.
It also means we can be honest about what we don't know. Public-data models are very good at reading form, but the market is sharp too. Ours does not beat the closing price — it beats the morning market, and the board then moves toward it through the day. So what we sell is the early price and how long it lasts, which is a smaller and more specific claim than most services make, and the only one our data supports.
What you actually receive
- Every selection of the day across UK and Irish racing.
- The registered price — win and place — on every selection, with the time we registered it.
- Morning delivery to the app, email and Telegram (paid tiers).
- An auditable record that grows every day — see the track record.
Start free. Upgrade when the model convinces you.
One selection a day free, plus yesterday's full card graded. The whole card is £14.99 a month.