ENGLISH FOOTBALL / MODEL EXPLORER
Every match. Every gameweek.
The numbers behind it.
Explore Premier League and Championship predictions for a whole gameweek or a single fixture. Compare 1 / X / 2 probabilities and color-coded outlook bars, then expand each match to see projected lineups and player ratings.
SEASON 2026/27 / PREDICT A ROUND
Gameweek predictions
01 / BUILD YOUR FIXTURE
Set the matchup
02 / READ THE OUTLOOK
Match probabilities
Every fixture has a story.
Choose two teams and a date.
Let the model put numbers to it.
Highest model probability, not a guaranteed result.
Prediction strength
Low: <10 pp / Medium: 10 to <25 pp / High: ≥25 pp. The gap measures probability separation, not proven accuracy or calibrated confidence.
How to read the draw probability
The draw percentage estimates the chance of any level score, not just 0-0. A draw can have a meaningful chance even when a home or away win is the most likely single outcome.
When enabled, experimental draw signals use low-scoring games (two total goals or fewer), goal margins and one-goal results from each team's last five matches, alongside how closely matched the teams are by Elo. These are learned inputs, not an automatic draw boost; they can raise or lower the estimate.
These additions have not established a reliable improvement in validation. Treat the probabilities as experimental estimates, not guarantees.
Starting lineups
Projected XIs are estimates from recent starters, not confirmed team news. Supplied XIs are labelled separately. Ratings are historical player-index values on a 3-10 scale, not predicted match ratings.
Inspect API response JSON
An experiment, not a crystal ball.
Explore a full gameweek from the local 2026/27 schedule, or supply a custom fixture and optional starting XIs. The model combines Elo, recent form, player ratings and league indicators, with experimental shot and draw-history signals when enabled. Projected lineups are not confirmed team news. Schedule dates may change, and future rounds use currently available history rather than simulated results. Probabilities are estimates, are not separately calibrated and do not guarantee an outcome.