We publish a number for every match we cover, show the market's number next to it, and grade both every night.
For each team we fit an attack strength and a defence strength on top of its own league's scoring rate, plus a term for playing at home, with recent matches counting for more and expected goals blended in where they exist. That gives two expected-goal figures for a fixture, and from those we derive the match result, over or under 2.5 goals, and both teams to score. The model refits twice a day.
A bookmaker's price is not a probability: the margin is built into it, so the three prices on a match add up to more than 100%. We take the margin out first, so a match page shows two probabilities in the same units rather than a number against a price.
Every night, every number we published is scored against the bookmakers' closing line — the sharpest public number there is. The record page shows how we did, market by market, including the markets where the close beats us.
When our number is at least four percentage points higher than the market's for an outcome we hold a price for, we mark it in purple as a spot. That is the whole rule. It is a disagreement, not a recommendation, and the bigger it looks the more likely it is that we are the ones who are wrong. A probability is not a tip. 18+, and please gamble responsibly.