A data company that happens to do racing

The problem we're fixing

The tipping world is built to look good in hindsight: winners are screenshotted, losing weeks quietly vanish, and "proofing" rarely means anything you can audit. Punters are asked to trust a personality. We think that's backwards.

Our approach

We built a machine-learning model on 38 years of racing — 3.8 million runner records across more than 400,000 races in the UK and Ireland. It rates every runner the same way every morning, with no ego and no narrative. Then we do the part the industry avoids: we register every tip before the off and grade it against the result, so the record speaks for itself.

What we sell is timing: the board moves toward our model between publish and the off, so the price on the morning card is usually better than the price at the off — and we have measured how long it stays available (92% of the time within fifteen minutes, 74% at an hour). We do not claim to beat the closing price; our own testing says we don't, and you won't find that claim on this site. We grade against it anyway, because it is the one number that can't be faked, and we publish the return alongside it — currently negative, and stated as such.

What we believe

• Transparent scoreboard, honest method. Show the process; earn the record.
• Probabilities, not personalities. The model speaks, we publish.
• Responsible by default. Tips are analysis, not a promise — and never a reason to bet more than you can afford.