Method

How the picks are made

The data

Every day we record the half-time and full-time score of every football match we can find — well over a thousand on a busy weekend, from top leagues to youth and regional divisions.

The model

For each fixture the model estimates how many goals to expect, starting from the country's average, then the league's, then each team's own games. Each step is weighted by how much data there is behind it, so a team with one match in the data barely moves the number and a team with twenty moves it a lot. The expected goals become a probability: the chance of at least one goal after half-time for the 2nd-half goal market, and of two or more goals in the match for over 1.5. The 20 highest chances are published each morning.

Why the record matters more than the chance

A chance of 88% is the model's estimate, not a promise. Before publishing anything we tested the model on past days it had not seen: its top 20 landed about 4 to 6 percentage points more often than a random game did, and it was a few points too confident. That is why every market page shows three numbers side by side: what landed, what the model predicted, and what a random pick would have done the same day.

Thin data

A pick marked thin data involves a team with fewer than three games in our records, so its chance rests mostly on league averages. As the records grow, fewer picks are thin and the chances get sharper.

Accumulators

If each of four picks has about an 85% chance, all four landing together has roughly a 52% chance — and in our tests the top four all landed only about two days in seven. Combining picks multiplies the risk, not the safety.

Bet responsibly

This site is for adults, it is free, and no bookmaker pays for anything on it. If you bet, bet small, and never to win back a loss.