Skip to content

Shot Lab

Drag the shooter, the keeper and the defenders. A model trained on thousands of real shots, each with every nearby player's position at the moment it was hit, tells you how often a shot like that goes in. This is machine learning: every weight in it was learned from those shots, none set by hand.

  • Shooter, with the ball
  • Keeper
  • Defender
  • Better spot nearby

How to read it

  • The goal is at the top; you're looking at the last third of the pitch, the penalty area in the middle.
  • The pale triangle runs from the ball to the two posts. Anyone standing in it is in the way.
  • The glow, with Heat on, is the chance from every spot if the shooter stood there and everyone else stayed put. Brighter is better.
  • The dashed ring is the best spot within about five metres, if one beats where the shooter is now. It assumes everyone else stands still, which in a real match they wouldn't.
  • Watch a move plays a short attack; the numbers change as the players move. Touch the pitch to take over.

Expected goals

–

Start from

Drag a player with your mouse, or pick one and use the arrow keys

How much the model knows

How did it decide?

What pushed this shot's chance up or down, compared with the average shot.

    How it works

    Logistic regression on open-play shots from StatsBomb's free data.

    Where it goes wrong

    • The freeze frame records the players near the ball, so a defender far away may simply be missing; the lab assumes the ones you place are all there are.
    • It doesn't know how hard or where the shot was hit, the shooter's body shape, or whether the ball was bouncing. Those decide a lot of goals.
    • A keeper off the line raises the xG here. In the data, a keeper far out is usually a keeper in trouble (rounded, or scrambling after a rebound), so the model can't tell you that coming out is a mistake.
    • Every shooter is an average shooter. A top striker scores more than this from the same spot; the model can't tell who's shooting.
    • Penalties, free kicks and corners are left out. The matches are from top leagues and tournaments, men's and women's, not the Scottish Premiership.

    How it was built and tested: Expected goals from scratch. Where everyone stood

    Want to see what a match's chances add up to? Put them into Did we deserve to win?

    StatsBomb

    Data: StatsBomb open data. The analysis is ours, not StatsBomb's. For education.