Click where the shot was taken. The model returns the probability that an average shot from there, in that situation, is a goal.

How to read the number

An xG of 0.30 means that, across thousands of shots taken from that spot in that situation, about three in ten were goals. It says nothing about the player. That is the point: it separates the quality of the chance from the quality of the finish.

What the model knows and what it ignores

It uses distance, the angle the goal mouth subtends from the shot, whether it was a header, whether it came from a set piece, whether the shooter was under pressure and whether it was hit first time. It does not know the goalkeeper's position, how many defenders were in the way, or who is shooting. Provider models that see those things will disagree with it, sometimes by a lot on individual shots.

The model, in full

Logistic regression fitted on 6347 non-penalty shots (570 goals) from StatsBomb open data, fitted on 2026-09-18. The coefficients below are the whole model; there is nothing hidden.

Feature	Coefficient
Intercept	-3,265947
`distance_m`	-0,149214
`log_distance`	0,932279
`angle_rad`	1,814928
`is_header`	-0,959428
`is_set_piece`	0,38758
`under_pressure`	-0,492055
`first_time`	0,007168

Calibration, predicted vs observed goal rate by bin: 0.00-0.05: 0,03 / 0,031 (n=2511); 0.05-0.10: 0,072 / 0,059 (n=2003); 0.10-0.20: 0,136 / 0,154 (n=1325); 0.20-0.30: 0,24 / 0,268 (n=272); 0.30-0.50: 0,374 / 0,37 (n=173); 0.50-1.00: 0,636 / 0,54 (n=63).

Data: StatsBomb open data (FIFA World Cup 2018, FIFA World Cup 2022, UEFA Euro 2020, UEFA Euro 2024, Copa America 2024). Log loss 0,2634, Brier score 0,0733.

xG calculator

How to read the number

What the model knows and what it ignores

The model, in full

Read the definitions