ANALYSIS

Expected Goals (xG) Explained: What It Measures and What It Misses

How xG is calculated, why it predicts future results better than actual goals, and the four situations where it lies.

Expected goals assigns every shot a probability of scoring based on distance, angle, body part, assist type and the defensive pressure around it. Add them up and you get a number describing how many goals a team should have scored from the chances it created.

It predicts future results better than actual goals because finishing is far more random than chance creation. A side that has out-created its opponents for six weeks while losing on the scoreboard is usually about to start winning, and the market is often slow to reprice that.

It lies in four situations: very small samples, matches decided by a red card, teams whose tactics genuinely produce better finishing positions than the model captures, and games where a side sat on a lead and stopped attacking. Treat xG as a strong prior that needs a human check, not as an answer.

Key points

xG measures chance quality, not luck or finishing
It predicts future goals better than past goals do
Game state distorts it heavily; leading teams stop creating
Six to ten matches is the minimum useful sample

Frequently asked

Is xG better than goals for predicting results?

Over any meaningful sample, yes. Finishing regresses toward the mean far faster than chance creation does.

What is a good xG per match?

Around 1.4-1.6 is typical for a strong side in a major league. Below 1.0 usually indicates genuine attacking problems.

Why do xG models disagree with each other?

Because they use different features and different training data. Directionally they agree; on individual matches they can differ by half a goal.

Put it to use

How our AI works →The open record →Today's picks →Free data API →

Expected Goals (xG) Explained: What It Measures and What It Misses

Key points

Frequently asked

Put it to use

Read next