The problem with save %

When was the last time you critically evaluated the metrics you use? Let’s do that now with save %.

Pros

• Intuitive: better goalkeepers should save more of the shots they face

• Improvement on goals against average: this doesn't consider the number of shots faced

Cons

• Treats all shots equally: the distance, angle, type of shot and many more factors will influence the likelihood of the shot being scored

• Analytics shows it is not repeatable year-to-year: this means it is heavily influenced by luck/randomness

The graph to the left shows that a goalkeepers save % in one season has almost no relationship with their save % the following season

Sample: NCAA D1 goalkeepers who played back-to-back seasons between 2011 and 2021.

We know some goalkeepers are better than others, and so should be able to repeat a certain level of performances, but if save % varies so much across seasons then it can’t be a good way of assessing goalkeeper performance. So if save % is not a good metric for evaluating goalkeepers, then what is? Below are my thoughts on some simple adjustments to improve your data collection.

Improvements

• Save % by zone: below the penalty spot vs. above the penalty spot.

• Save % by situation: 1v1s, set defense vs. counter attack, field vs. penalty corners etc.

• Save % by goal location: glove side shots vs. stick side shots,