How it works

We start from the bookmakers' own price with their margin removed. Our model only moves away from it where the data gives a real reason, and only as far as its own uncertainty allows.

One simulation, every market

Each match is simulated 10,000 times. The result, goals, both teams to score, corners, cards, fouls and shots are all read from the same simulated games, so they never contradict each other. Referee strictness, derbies, table stakes, rest days and missing regular starters all feed in.

Plain-words reasons

Every reason you read is built from the numbers: we switch each factor off, recompute, and report how many points it moved. If the words and the numbers ever disagreed, that would be a bug.

News changes picks, and we show it

When lineups, injuries, the referee, the kickoff time or the odds change, the match is re-predicted within minutes. A new version is published only when something material moved, and it says what changed.

A ledger nobody can edit

Each pick is hashed into a public chain before kickoff and settled after the match, with the closing price. Check any entry.

When our data feed fails

If our data provider stops answering, or our model hits an error, we publish the bookmakers' price as it is, tagged Avoid, and say so in a banner. A model running on stale data could look confident and be wrong.

What we do not claim

Closing prices in the biggest leagues are very efficient; beating them on match results is rare. Our edge is more often in corners, cards and fouls, and in reacting fast to team news. Bet only what you can afford to lose.