The price, before
the market moves.
A model trained on 38 years of racing rates every runner each morning. We publish the selections with the price we registered — and we have measured how long that price lasts: still there for 92% of members after fifteen minutes, 74% after an hour. Every selection is graded in public afterwards, return included.
The dataset behind every selection — 1988 to today, refreshed daily
Today's free pick is live
The model publishes a free selection every morning — registered at its price before the off and graded tonight like everything else, win or lose. Create a free account to see today's and get it delivered every morning. The record so far →
The model's work, in your account
Every morning: every selection of the day on both bases — win and place — with the price we registered, how long it lasted, and the model's read of every runner. Click through a preview of the member experience.
A model, not a man with a tip
No insider whispers, no gut feel. Every selection comes from a machine-learning model trained on decades of form, registered before the off and graded the same way every day — whether it flatters us or not.
Decades of racing
38 years of results, form, ratings, going, draw and connections — refreshed daily across UK and Irish racing.
Probabilities, scored
A gradient-boosted model rates each runner's chance of placing. No cherry-picking — the model speaks, we publish.
Registered & graded
Every tip is logged at the price available, before the off, and graded against the result — so as our record builds, we can publish it in full.
Four decades of racing, in one model
Most tipsters sell a hunch. We built a machine that has studied more racing than any person could in a lifetime — and reads every card the same way, every day.
The dataset
Form, ratings, going, draw, class, connections and prices — cleaned, joined and refreshed daily across UK and Irish racing.
How the model thinks
| Learn | Gradient-boosted win & place models on 38 years of results |
| Calibrate | Probabilities corrected against outcomes, then checked against the live record |
| Validate | Walk-forward tested on unseen seasons, strict leakage controls |
| Publish | Each morning's tips registered at the price available, pre-off |
What we actually sell: the early price
We don't beat the closing price and we don't claim to. What we do is publish before the market moves. Across 13,202 live runners, the horses our model rates highly hold or shorten while the ones it rates poorly drift out — and the price we register is still available to 92% of members within fifteen minutes, 74% at an hour. It goes fastest on exactly the selections that end up beating the closing price. Every tip is registered before the off, graded automatically, and published with its return, losers included. How we measure it →
The model's card, every morning
Every selection the model makes that day, with the registered price — win and place — on each one, delivered to the app, email and Telegram before the first race.
Six things you can check yourself
Every line here links to the page that proves it. None of it asks you to take our word.
- Gradient-boosted win and place models on 38 years of form — 3.8M runners across 400,000+ races. The same process every morning.
- Logged with its price in a ledger that is never edited or deleted, and each morning's full card is SHA-256 committed before racing so it can be checked afterwards.
- The record shows every settled selection, placed or not. Nothing is removed after the fact.
- Every settled selection with its registered price, starting price and result — as a CSV, no account needed. We don't headline a return figure, because it depends on your stakes and the prices you take; this is everything one would be computed from.
- 92% of registered prices are still available fifteen minutes later, 74% at an hour — and they go fastest on the selections that beat the close.
- Betting carries risk and variance. What we promise is an auditable process, not an outcome.
One plan. £14.99 a month.
We used to sell three tiers, and the upgrade between them bought more selections. That is the wrong thing to charge for, so there is one plan now — and the record is on the site before you decide.
- Yesterday's full card, graded — every day
- One of today's selections, with its registered price
- The published record in full, return included
- The methodology, the data studies and every calculator
- Every selection of the day — win and place, UK & Irish racing
- The registered price the moment it publishes, with how long it lasted
- Live fair-price check on every pick — our price against the best board price
- Price alerts the moment a runner crosses our fair line
- Tomorrow's card the night before
- Bet tracker, staking tools, API and full data export
Cancel any time from your account. No lock-in, and a statutory 14-day right to cancel for a refund.
The honest answers
Will I definitely make money?
No — and anyone promising that is lying. Betting carries risk and variance. What we promise is transparency: a model-driven process, with every tip registered and graded so that — as our record builds — you can judge the edge for yourself.
What am I actually paying for?
The whole card the moment it publishes, on both bases — win and place — with the price we registered and how long that price stayed available. Plus the price alerts, the night-before card, the tracker and the tools. What you are not paying for is a promise about winners: our record is published in full, including the return, and you can read it before you decide.
How are the selections produced?
A machine-learning model rates every runner each morning from decades of data — no tipsters, no opinions. Selections are registered before the off and graded against the result automatically.
Can I cancel anytime?
Yes — cancel whenever from your account, with no lock-in. Because you buy online, you also have a statutory 14-day right to cancel a new subscription for a refund; see our Terms for how that works alongside the free trial.
Data studies, straight to your inbox
What 38 years of racing actually tells us — CLV, biases, value. No spam, unsubscribe anytime.