Goalkeeper statistics have been left behind in football's data revolution…
Clean sheets are still used to decide Golden Glove awards when they have been shown to measure defensive quality rather than goalkeeping quality. But it isn’t just clean sheets; the majority of commonly used goalkeeper statistics do not measure inherent goalkeeper ability when put under the spotlight.
Most ‘Team of the Season’ awards are based on vague performance averages collected over the course of a season. Goalkeeper selections are not based on data that has drawn on goalkeeper-specific variables.
Goals conceded is strongly correlated to shots faced meaning that year in, year out the goalkeeper who concedes the most goals is likely just the goalkeeper playing behind the worst defence. Likewise, the goalkeeper who conceded the fewest goals is likely just playing behind the best defence. This doesn't necessarily mean the goalkeeper has or hasn't performed well, but these stats don't isolate their specific actions, which is what is necessary to determine the goalkeeper's individual contribution to a given match or season.
Save percentage is also frequently used to analyse shot stopping ability. However, it inherently treats every shot equally (in terms of ‘difficulty’) which, over the course of a season, has been shown not to even itself out. No shots are the same and they should be treated that way. The individual characteristics of every shot should be taken into account when evaluating each shot’s save difficulty.
Errors Leading to Goals are also often cited as a reason why a goalkeeper is exceptional or poor but it is a subjective measure which different observers would measure differently based on their perception of what qualifies as a mistake. No usable analytical statistic can be subjective.
Outside Penalty Area Actions are often used to measure a goalkeeper's sweeping ability, but not all actions outside the area prevent a chance occurring while some actions inside the area prevent huge 1v1 chances occurring. Ignoring data points which evaluate a skill you are attempting to measure is a big mistake when analysing a sample.
Because of these discrepancies, there has often been clashes between the goalkeeping analysis which occurs inside of football clubs, which mainly relies on the ‘eye test’, and the goalkeeping analysis which occurs in the media. At Goalkeeper xG we are revolutionising goalkeeper statistics and data in order to close this divide.
So, what do we do differently?
At Goalkeeper xG, we model not just every action a goalkeeper performs, but every action a goalkeeper could possibly perform using sensible methods from a goalkeeper first perspective.
This allows us to quantify the difficulty and relative importance of everything a goalkeeper does on the football pitch. We finally answer elusive questions such as ‘should the goalkeeper have saved that?’ or ‘was the goalkeeper right to come off their line?’, and ‘does that goalkeeper’s good sweeping and distribution outweigh their average shot stopping?’.
Goalkeeper xG's models are the most goalkeeper-specific, and therefore most accurate, measures of goalkeeper performance available in the sports analytics industry today.
Over the course of the 2022/23 Premier League season, we have collated our own End of Season Awards. Our models have analysed every single potential goalkeeping action in the Premier League and have found, quantitatively, which goalkeepers made the biggest difference to their clubs in various categories.
Stay tuned over the next two weeks to see the real End of Season Goalkeeper Awards…