STATS Score Index Methodology
To create league rankings for goalkeepers, we take all shots faced by all goalkeepers to date in their league. Traditionally goalkeepers are assessed using metrics such as clean sheets, goals conceded and save percentage. More recently “expected” metrics such as expected saves (xS) have been introduced to compare performance to the league average. However, goalkeepers could have completely different types of saves to make depending on the defensive style of their team and the opponents they face.
A set of embeddings captures the different skillsets of a goalkeeper, which we can then swap in order to simulate their performance. This gives use the total number of goals each keeper would have been expected to concede (xG) if they faced every shot in a given sample. To make this more interpretable, we first standardise the expected goals from a season total to per game, then take this from the league average. This final value is the number of expected goals a goalkeeper would have either prevented or conceded compared to the league average.
For example, if we look at the English Premier League rankings for the first week in April 2019, Liverpool’s Alisson and Manchester United’s David De Gea are expected to save 0.33 more goals than the league average. At the other end, Bournemouth’s Asmir Begovic is expected to concede 0.32 more goals than average. This means, simulated against a sample of 2,708 shots from the 2018/19 season, De Gea and Alisson would be expected to save an additional 0.65 xG per 90 minutes versus Begovic. Finally, we only show goalkeepers’ who have played in 10 or more league games.
Our novel method can be used by coaches and recruitment departments to better analyse keepers and understand their strengths and weaknesses. If you’re interested in finding out more about the method, follow this link to Trading Places – Simulating Goalkeeper Performance using Spatial & Body Pose Data.