Expected Goals

Expected goals, goals against and points

How far to trust the xG values · Premier League 2026/27

Does a shot with an xG of 10 % really end in a goal about one time in ten? Methodology

xG calibrationShare of shots that were goals against the mean xG of the shots, per xG bin. The closer a point is to the diagonal, the better the xG model matches reality.0%0%20%20%40%40%60%60%80%80%Mean xG of the shotsShare that were goalsxG 2.89742852715%: 2.96643247463% goals from 1281 shotsxG 7.0257181197%: 6.36094674556% goals from 676 shotsxG 12.3517503934%: 10.0746268657% goals from 268 shotsxG 17.168223502%: 14.9253731343% goals from 134 shotsxG 22.6420979251%: 18.6813186813% goals from 91 shotsxG 27.6024314918%: 28.0701754386% goals from 57 shotsxG 32.8531499505%: 26% goals from 50 shotsxG 37.2901546458%: 33.3333333333% goals from 24 shotsxG 42.0745761267%: 39.2857142857% goals from 28 shotsxG 47.7946106683%: 36.3636363636% goals from 22 shotsxG 66.3536057796%: 51.8518518519% goals from 81 shots
Perfect calibrationShare that were goalslarger dots · more shots

Calibration error 1.39 pp, goals per xG 0.87 across 2,712 shots.

Shots by xG
xG of the shotShotsMean xGGoals
0.00 – 0.0512812.9%3.0%
0.05 – 0.106767.0%6.4%
0.10 – 0.1526812.4%10.1%
0.15 – 0.2013417.2%14.9%
0.20 – 0.259122.6%18.7%
0.25 – 0.305727.6%28.1%
0.30 – 0.355032.9%26.0%
0.35 – 0.402437.3%33.3%
0.40 – 0.452842.1%39.3%
0.45 – 0.502247.8%36.4%
0.50 and up8166.4%51.9%

The shots are sorted by their xG. If the xG values are right, a group of shots with a mean xG of 10 % ends in a goal about one time in ten. The calibration error is the average gap between the two, weighted by the number of shots. Penalties are left out. xG models differ between providers, see the methodology.