We started as a tips blog in 2015. It's now something rather different: a ratings engine and a betting model that price every race, every day, automatically.
It's been a fun ride since we started as a tips blog in 2015. It was always a hobby and, candidly, we can't compete with professional tipping services and media outlets on volume of output — the most loyal of you will have noticed ours dipping in recent years, with tipping time at a premium.
Even back at the beginning I had visions of the website being more dynamic and automated, but I couldn't write a line of code. A lot has changed since then, and now I don't have to. What started as a pet project to automate my betting took me down a long and fascinating rabbit hole through racing APIs, machine learning and AI platforms.
I'd been running my own betting algorithm behind the scenes since November 2025, but that came to an abrupt halt earlier this summer when Racing Post Ratings — crucial for quantifying form and making predictions — disappeared from my data provider. That left me with a choice: give up, or build my own ratings system from the data I already had.
That's how 2PRs were born — every horse performance rated across GB, Irish, French and UAE flat racing, plus GB and Irish jumps, with every horse in the population ranked across distances, codes and ground conditions. You can browse the whole population by division, look up any horse's profile, and even build a hypothetical field in the Race Simulator and have the model price it up. These figures are designed to give me exactly what I need to feed my machine learning algorithm, The Feckalyser, whose sole job is to price markets and identify value bets.
I've loaded a freemium version of The Feckalyser onto the site. It posts its three best bets at 11am every day, here and on X; the rest of the day's qualifiers go to free members by email — no strings attached. For now it tips flat racing only. The jumps engine is under construction and I'm aiming to have it ready for the jumps season — the 2PR jumps ratings are already live in beta while that work goes on.
Be patient with it. A positive expectation only shows up over a long run of bets, and the same simulation contains a maximum drawdown of −61.0pts — long losing runs are table stakes for any betting model with a comparable ROI. A backtest is a measurement of the past, not a promise about the future.
The sharpest, highest-resolution and most productive version of The Feckalyser is reserved for me for now, as it leans on live Betfair Exchange API data which I'm only permitted to use for personal betting. I hope to share it with you in future if it receives their blessing.
So that's every in-scope race analysed automatically, every day, with automated tipping built on top — but Paul and I will still be posting our house tips, focusing on weekend racing and the bigger festivals. You can even watch us against the machine: the record page has a man-vs-machine view comparing human judgement with the model, on the same book and the same rules.
Every single bet is tracked and settled against a public record, fully transparent — winners and losers, nothing quietly deleted. No amount of backtesting can guarantee the future, but if you join me for the journey we'll win and lose together.
Viva twopointswin!