Expected Goals (xG) Explained: How to Test It Against Results
Expected goals (xG) estimates the scoring probability of a shot by comparing it with similar chances in a model’s historical data. A shot rated 0.2 xG represents an estimated 20% chance in that model’s reference cases—not a promise that the player will score one goal from five attempts.
What xG can and cannot tell you
A model may use features such as shot location, angle, body part, and situation. Providers can define models differently, so two xG feeds may not match. Team xG adds the estimated values of its shots. It helps describe chance creation, but it does not capture every tactical action or fully explain game state, passing before the shot, defensive pressure, or keeper positioning.
Design a small historical test
- Choose a competition, season range, and a trusted dataset with documented xG definitions.
- Keep every match in the selected sample, not only famous or high-scoring games.
- Record each team’s xG and goals, then calculate the difference.
- Compare over many matches: average error, calibration by probability bands, and how often the higher-xG team won.
- Split data into development and later test periods if you tune any rule.
Read the results with care
If a 0.2 chance is repeated many times, outcomes should approach the model’s average, but short runs remain noisy. Report the sample size, missing matches, provider, and method. Don’t use a chart from a small sample to claim xG predicts a specific fixture or guarantees a betting edge.
Build a useful visualization
Start with a scatter plot of team xG against actual goals and draw the equal line. Points above it scored more than their xG in that sample; points below scored less. Add one dot per team-match, a clear axis label, and a note that the provider’s model is an estimate. Avoid ranking players by xG alone without minutes, role, and shot volume.
Try it: select ten historical matches and calculate the difference between xG and goals. Then expand to a whole season before drawing conclusions.
Keep learning: apply the same careful approach to T20 scoring trends →
Definitions and data context: Premier League xG glossary and Premier League statistics clarification.