Data for Dummies #1
xG, xGA and xPTS - explained.
Data for Dummies is a new series coming to More Than A Game, with each article accompanied by a video broadcast on our YouTube channel.
Some say there is only one set of numbers that matters in football – the scoreline.
However, at More Than A Game we like to delve into the data to bring you closer to the biggest stories, to explain exactly what a player offers or why something is happening on the pitch.
After all, if the world's most successful clubs consult data when recruiting players and coaches, why shouldn't fans hold an interest in numbers?
With data becoming increasingly visible in the football world, we compiled this handy explainer to tackle some of the more ambiguous terms being thrown around.
In the first article of a new series, More Than A Game dives into expected goals, expected goals against and expected points.
What is expected goals (xG)?
Loved by some, loathed by others and misunderstood by many, expected goals is one of the most visible metrics in football.
It measures the quality of a chance, assigning a likelihood of it being scored – an xG value between zero and one — based on several factors, including shot position and angle, the type of shot (e.g., with the head or foot), type of assist (e.g., cross, cutback etc.) and pattern of play. Each attempt in a game is given an xG value between 0.01 and 0.99 (or between 1% and 99%).
Opta — who first introduced xG in 2012 — say their model is "powered by hundreds of thousands of shots from historical data". In a nutshell, looking at whether certain types of chances were historically converted tells us how likely they are to result in a goal.
Players and teams that post goal returns above their xG figures over a given period are said to be overperforming, those who do the opposite are underperforming.
However, when it comes to a team’s xG, it is not always a positive to overperform, as that overperformance is unlikely to be sustainable throughout an entire season.
When it comes to a player’s xG, then an overperformer will be somebody who displays extremely clinical finishing, above what would be expected of the ‘average’ player. For example, Erling Haaland scored 36 league goals last season, from an xG of 28.6.
We can delve into more detail in future editions of this series, but essentially xG is an excellent way of measuring how a team or player are performing from an attacking perspective. If a team creates more chances, they are likely to have a higher xG than their opponents.
Of course, football is not played on paper, or on a spreadsheet, so while a better xG should, in theory, lead to better results, it is not always the case. See, for example, Everton so far in 2023-24.
As per Understat, Everton have accumulated 7.95 xG across their four Premier League games, which ranks Sean Dyche’s side 10th in the competition. For ease, let’s round that up to a round 8.0. However, they have only scored twice, meaning they are underperforming by 6.0 (six) goals. Given the lack of quality up top for Everton, the numbers are — in this case — backing up the eye test.
What is expected goals against (xGA)?
This is easy. xGA is simply a way of flipping the expected goals metric from quantifying attacking quality to defensive.
If a team’s xGA is low, it generally means they are performing well defensively by limiting the chances of the opposition. If it is high, then they are allowing plenty of opportunities for their opponents to score.
Many observers and pundits would still say the team that concedes the fewest goals across any given amount of time — let’s say a 38-game league season — has “the best defence”. But while that has an element of truth to it, there are other factors that you have to consider when a team concedes the fewest goals. Was that all down to the quality of their defensive play, or did they simply get lucky due to bad misses from the opposing attackers, or did they have a goalkeeper who was on top form?
Of course, “the best defence” in terms of goals conceded could also match up to the best defence when it comes to xGA. Indeed, the two would in theory go hand-in-hand across a campaign, just as you would expect the team that creates the highest xG to score the most goals. But it is not always the case.
Let’s use last season as an example.
Premier League champions Manchester City recorded an xGA of just 34.21 and only conceded 33 goals. That shows that the defence was performing extremely well, and also poor finishing or quality goalkeeping will have played its part, but only to the tune of one goal fewer than City were anticipated to concede based on the statistics.
Wait, though. Newcastle United also only conceded 33 times in their 38 Premier League matches. However, in Newcastle’s case, their 41.86 xGA was noticeably higher than City’s. While the Magpies still had the second-best xGA figure in the league, they actually overperformed by 8.86.
Like an overperforming attack, an overperforming defence is not always a positive when it comes to sustainability. Newcastle benefitted from some poor finishing and excellent goalkeeping from Nick Pope last season, but so far this time around — and it is a small sample size — they have not been so lucky, with Eddie Howe’s team conceding 2.11 more goals than would have been expected based on the quality of chances and shots their opponents have created.
What is expected points (xPTS)?
Expected points is a calculation of the xG and xGA from any given game.
In a basic model, the xPTS is calculated by collecting all the shots within a game, with each shot given an xG value of between 0.01 and 0.99 (or 1% through to 99%).
A model will then run this through hundreds, if not thousands, of simulations to produce an xPTS number based on the averages of those results.
Soccerment describes xPTS as such:
The fraction of times the outcome is positive, i.e. a goal is predicted, is the probability of the shot resulting in a goal. By applying this method to all shots occurring in a match, we can obtain the probability for each team to win, draw or lose. Expected Points are then simply the sum of the products between each probability and the corresponding points.
So, what’s the upshot?
Well, it means that fans and data nerds alike can assess where a team should be placed in a league based on the quality of chances they both create and concede.
Back to Everton, and via Understat, the Toffees are currently underperforming their xPTS by 5.18. Based on the quality of Everton’s play at both ends of the field, they should have taken six points from their opening four league matches, yet instead of finding themselves in mid-table, they are in the bottom three having taken only one point. That differential of 5.18 is the biggest in the division, meaning there’s a fair argument to say Dyche’s side have been one of the most unfortunate teams when it comes to not getting what they deserve.
They are followed in that regard by Chelsea, with the Blues having performed well enough to take 8.15 (eight) points, yet they only have four to show for their efforts.
City lead the way in reality, and so they do in the xPTS table, albeit they have taken two more points than the model suggests they should have done.
Burnley are bottom in both tables, with 0 points in reality and only 1.17 xPTS, though it must be said the Clarets have only played three games.
By Patric Ridge