Horse racing, greyhounds and football — every runner priced by model, then measured against the market. Four systems, one feature engine, one standard of proof.
Three steps, in that order, across every sport we model. No tipping on a feeling, no quietly forgetting the losers, no results that only exist as screenshots.
In a horse race: around two hundred features per runner — ratings, time-decayed form by surface, course, going and distance, pedigree, headgear, field strength. In a greyhound race: the pace map — split to the first bend, running line, trap fit, traffic. Ensembles turn it into one number.
The model’s price is set against the live market. A bet is only issued where our probability is materially higher than the odds imply — a genuine overlay, not a rounding error. Stakes are sized by Kelly, at whatever fraction you set.
Every selection is settled three ways: at the odds we advised, at Betfair SP and at Industry SP. Rule 4 deductions applied, voids marked, nothing edited after the fact. The ledger below is the whole record.
One feature engine shared by training and live scoring, and an audit that proves the model only ever saw what it could have known on the day.
Around two hundred features per runner, blended into a single price.
Pace modelled from the run to the first bend, where the race is decided.
Eight boxes, two grading ladders, every meeting priced on the exchange.
Define conditions and a market; get hit rate and ROI, season by season.
Strip the bookmaker’s margin out of a set of odds and something clean is left behind: a list of probabilities that sums to exactly one. Every runner is a coordinate. Every race is a single point inside a simplex.
The market’s point and ours rarely sit in the same place. Most of the time the gap is noise and we do nothing. Occasionally it is structural — the market has mispriced a runner, and the distance between the two points is the edge we are paid for.
That distance is measurable before the race and verifiable after it. Closing line value tells us whether we were on the right side of the move; profit tells us whether it mattered. We publish both.
Built for people who already know what they’re doing with a price.
Full horse racing cards with model probability, market comparison and expected value on every runner — not just the ones we bet.
Tips issued only where the model finds a real overlay, win or each-way. Some days that is four bets. Some days it is none.
P&L, ROI and CLV at advised odds, BSP and ISP. Rule 4s applied, voids counted. The bad months stay on the chart.
Stakes sized to the edge, with your own Kelly fraction and bankroll. Most people run a quarter or a half. Nobody sensible runs full.
Combine the day’s value picks into multiples with live combined odds and slip management. Compounding edge, compounding variance.
Slice the card by class, going, distance, trainer form or model confidence, and keep the views that earn their place.
Every settled selection, priced three ways. Losing days are on the chart in the same colour as the winning ones.
| Date | Bets | Won | Placed | Lost | Staked | P&L | ROI |
|---|---|---|---|---|---|---|---|
| Loading the ledger… | |||||||
The mathematics underneath the product, written up properly — convex optimisation, the geometry of a betting market, and what the numbers actually support.
Live racecards, model prices, value selections and the full ledger — all inside the portal.
Launch the portal