Expected goals, or xG, is the probability that a shot taken will result in a goal. The xG value of a shot can be between 0 and 1, representing that probability. A shot with an xG of 1 will always go in; a shot with an xG value of 0 will always miss. Different companies and sites have different calculations and models for their xG values, but most use similar variables to determine them: the distance and angle of the shooter’s location compared to the goal, the type of shot (a header or from the foot), the number of defenders between the shooter and the goal – plus more. The shot in question is compared to hundreds or thousands of shots with similar characteristics to get a probability that it will go in.
Often, xG values will be presented as a total of all the shots taken in a game and will suggest how many goals a team should have scored based off their chances. A team with a high xG over a game is often better at creating – Liverpool (89.2), Manchester City (89.0) and Bayern Munich (88.1) created the most xG in Europe’s top five leagues this season. A penalty has an expected goals value of around 0.76 depending on the model being used. npxG, or non-penalty expected goals, takes away penalties from a team’s xG total.
Expected goals can also be used to judge a player’s individual performance. A player with a high xG over a game is often one that is good at finding dangerous positions in attack – Robert Lewandowski, who on average had the highest xG per 90 minutes in Europe (1.00), would be expected to score 1 goal every game. This is in part due to his movement, but also partially helped by the creativity of his Bayern teammates. Lewandowski, however, scored an average of 1.07 goals a game. The fact that this figure is greater than his xG per 90 suggests that he is scoring more than expected – a sign of a clinical finisher.
xG also has other uses. Whilst it can be used as a tool to judge a striker, it can also be used to critique a team’s defence by looking at the amount of xG a team concedes. Per 90 minutes, Norwich City conceded an average of 2.03 xG – in every game they played last season, their opponents would have created enough chances to be expected to score 2.03 goals. In contrast, Napoli’s opponents would only be expected to score 0.82 per game. A team with a lower xGA (expected goals against) value is a lot better at preventing goal scoring opportunities, whilst a team with a higher xGA is caught out more often (which isn’t much of a surprise given Norwich’s terrible defensive record).
Another, more niche use of expected goals is the evaluation of a goalkeeper’s shot stopping ability. For this, post-shot xG (PSxG) is used. Regular xG takes several variables into account, but this is before the shot is taken. What xG doesn’t account for is the quality of the shot once it’s been taken. For example, a scuffed shot from 30 yards out that goes wide and a perfectly placed free kick from 30 yards out may have the same xG – but post-shot xG would tell a different story. Off target shots have a PSxG of 0 – a shot that goes off target cannot result in a goal. Shots that have a higher PSxG than xG will be shots that have been hit well. For a goalkeeper, we can use PSxG to take away off target shots and look at how they have dealt with the shots they’ve faced.
Mike Maignan is currently celebrating AC Milan’s first Scudetto in a decade. Over the course of the Serie A season, he faced 28 PSxG worth of shots. That means that he would have been expected to concede 28 goals. However, he only conceded 21. The difference here is +7, meaning that effectively, Maignan prevented 7 goals this season, or 0.21 goals in each of his 32 games. That +7 number is his Post Shot xG minus Goals Allowed (PSxG – GA) figure for the season. A positive PSxG – GA figure suggests that a goalkeeper has shown good shot stopping ability, preventing goals from being scored. Players like David de Gea and Jose Sa have largely positive PSxG – GA values.
In Real Madrid’s Champions League victory over Liverpool, Thibaut Courtois kept a clean sheet, but had PSxG – GA value of +3.5 – that is, an average goalkeeper would have conceded 3.5 goals given the shots Courtois faced. Numbers like this really quantify and highlight spectacular goalkeeping performances like this. It can, however, also bring to light some harsher truths. Aaron Ramsdale, who has often pulled out great saves for Arsenal this season, had a value of -5.6 over the course of the season. A negative number suggests that a goalkeeper is letting in more than they should. Of course, there are other variables at play, including luck and the fact that Ramsdale had only just taken the step up from Sheffield United. Stats should always be used in context, but a keeper’s shot stopping ability is not something to ignore.
Expected goals have a huge part to play in football analytics, quantifying things like a striker’s ability to finish or a goalkeeper’s ability to stop shots. Metrics like expected threat (xT) are now coming into use and were partially derived from xG. Pundits on Sky Sports will now mention xG in their post-game analysis. Arguably, xG is what gave football analytics its first bit of publicity in the mainstream. However, with more refined models being made all the time to improve accuracy and precision, xG isn’t something that will be going away any time soon.
