What is GSAx? Goalkeeper Analytics Explained Simply

How do you measure a goalkeeper's quality? Clean sheets? Save percentage? Goals conceded? All of these are deeply flawed because they don't account for the quality of shots the goalkeeper faces.

A goalkeeper behind a dominant defense might face only 2 shots per match, all from long range. Another behind a porous defense might face 8 shots per match, several from inside the six-yard box. Comparing their clean sheet records tells you more about the defense than the goalkeepers.

This is where GSAx comes in. It isolates goalkeeper performance from defensive quality.

GSAx stands for Goals Saved Above Expected. It compares the number of goals a goalkeeper actually concedes against how many they "should have" conceded based on the quality of shots they faced.

Here's the calculation: for every shot on target a goalkeeper faces, the xG model assigns a probability of it being scored (e.g., 0.35 for a shot from 12 yards at a moderate angle). If the goalkeeper saves it, they get +0.35 GSAx for that shot. If the shot goes in, they get -0.65 GSAx (they failed to save a shot that goes in 35% of the time, meaning 65% of the time a save would be expected).

In a typical Premier League season, the best goalkeepers achieve GSAx values of +5 to +8, meaning they save 5-8 more goals than an average keeper would. The worst performers might sit at -3 to -5.

It's worth noting that GSAx can fluctuate significantly over small sample sizes. A goalkeeper might have a GSAx of +3 after 10 matches, then regress to +1 after 30. Season-long GSAx is much more reliable than match-by-match values.

Save percentage treats all shots equally — saving a shot from 30 yards counts the same as saving a point-blank header. GSAx weights saves by difficulty, giving more credit for stopping hard chances.

A goalkeeper with a 70% save percentage might have a negative GSAx if most of their saves were easy (low xG shots), while the goals they conceded were from relatively saveable positions. Conversely, a 65% save percentage goalkeeper might have a positive GSAx if they're stopping shots that most keepers would concede.

At xGoal, GSAx is a key input to the Goalkeeper signal. The engine combines it with other goalkeeper metrics to assess how much each team's keeper adds to (or subtracts from) their defensive capability.

The goalkeeper signal works alongside the defensive reliability signal. A team might have a weak defense (high xGA) but an elite goalkeeper (high GSAx), meaning the actual goals conceded are fewer than the defensive metrics would suggest. This nuance is lost if you only look at one metric.