xG Lab

The transparent football model

Football predictions with nothing hidden.

xG Lab is an open-logic football analytics project. Every win probability, projection and rating comes from a model we explain in full — then we grade it honestly against the market consensus, wins and misses alike. No black box. No hype. No paywalled picks.

Open logic, no black box

Every probability traces back to a published model: rolling xG-based team strength → scoring rates → a minute-by-minute Poisson engine. The glossary explains each metric in plain English.

Graded against the consensus

We line our pre-match probabilities up against the closing market consensus and show how close we get — including where the consensus beats us. Honest accuracy, not cherry-picked wins.

Free and independent

Informational analysis for football and data fans. No ads-driven hype, no paywalled picks, no affiliates — just the model and the numbers.

What you get

How the models work

Win probabilities come from a rolling team-strength model: each team’s attack and defence strength is estimated from recent xG (an EWMA that carries across seasons), turned into pre-match scoring rates (λ), and fed into a Poisson in-game model that updates minute by minute. Season projections are Monte Carlo simulations of the remaining fixtures from the same strength estimates.

New to the metrics? The glossary explains xG, xGChain, PPDA, expected points and the prediction model in plain English.

Disclaimer

All probabilities, projections, ratings and derived metrics are statistical estimates provided for informational and educational purposes only. They are not advice of any kind, and no outcome is guaranteed. Third-party data may contain errors or delays. Use the information at your own risk.

Data sources & attribution

  • Match results, fixtures, standings and lineups: football-data.org.
  • Expected goals (xG), shots and advanced player and team metrics come from a licensed commercial data provider; win-probability, projections and ratings are our own calculations on top of it.
  • Event data used for passing networks: StatsBomb Open Data (CC BY-NC-SA), used with attribution for non-commercial illustration.
  • Player photos: TheSportsDB. Country flags: flagcdn.
  • Club names, crests and logos are trademarks of their respective clubs and are used for identification only.

We store only headlines, short summaries and links for news items and link back to the original publishers; we do not reproduce full articles.