Even well before the advent of widely available AI, the role of stats and data in football had grown at a rapid pace in the 21st century. Gone are the days when players are bought at the say-so of a wily old scout who can spot raw talent. Data is crucial in just about every area of the game nowadays, be that in terms of dynamic pricing for hospitality packages, how much to pay a player and what length of contract they should have, or in tailoring the recovery of an injured footballer.
Numbers also now play a huge part in how pundits and social-media experts analyse the game and how viewers and fans interact with it and assess it too. There are more stats in the game than ever, with metrics to assess and quantify just about every aspect of football, individual matches and player performances.
One of the biggest and most widely used stats is xG, or expected goals. It may surprise many that the term is thought to date back as far as the early 1990s, when researchers were looking at the impact of artificial pitches. Back then, it was used in a slightly different fashion than how it is used now, and the history of the idea of expected goals is a little disputed, evolving over time and also having roots in ice hockey.
Leaving aside the origins of xG, in its modern guise, we can date its usage in the Premier League to 2012/13, with Sam Green, at Opta, sometimes credited with its “invention”. The official PL website states that when the competition began, they only collected a “basic level of match data”, but that “since 2006/07, a wide range of statistics” are available. They list when each metric became available, and while stats like sweeper clearances by goalies were added in 2006/07, xG did not enter the fray until 2012/13.
What is xG?
The expected goals stat is now understood by most fans, even if many remain a little ambivalent and a smaller number strongly question its merit. For some, it is as mainstream a stat as possession or shots on goal, but for others, it remains a concept they don’t fully understand.
The simplest way to think about xG is as a stat that shows the quality of the chances a side had in a game. Shots and shots on target were the traditional way to understand who had the better of a game, alongside possession. However, basic shot stats do not differentiate between an effort from 40 yards that bobbles slowly to the goalkeeper waiting in the centre of the goal and a free header from the six-yard box that is planted into the corner.
Of course, ultimately, as commentators are wont to say, the only numbers that matter are the ones that give the score. But at the same time, that does not mean other stats are pointless. Critics of xG question various aspects of the stat, but that misses the essential point that it should simply be viewed as one tool that can be used to obtain a better understanding of the game.
Chance of Scoring Measured
It is not perfect, but it provides a clearer picture than shots alone. That is because it effectively gives every shot a mark out of 100 according to the quality of the chance. It is not really a score, but actually a probabilistic assessment of the chances of such an opening leading to a goal.
A shot that has almost no chance of going in will have an xG of 0.00. This is because xG is calculated using two decimal places, such that a chance that is deemed to have an xG of less than 0.005 will be given zero. The highest value given is 1.00, but this is uncommon, equating to a chance that would be converted as close to every time as possible.
In simpler terms, an xG of 0.01 means that statistically speaking, such an effort would only go in one time in 100, whereas an xG of 0.99 would be expected to lead to a goal 99 times out of 100. Various aspects of the chance are factored into what is a very complex equation to derive the xG.
How is xG Calculated?
There are different xG models, and so you may sometimes see different scores, but Opta’s is the most widely recognised. They have analysed historical data, which is to say hundreds and hundreds of thousands of shots. The initial training data for the model analysed shots from 40 separate competitions between 2018/19 and 2021/22. From that, they extrapolated how different factors affect the odds of different chances leading to goals.
Obvious factors include the distance from goal and the angle, but some xG models also take into account what type of shot it was (header, volley, or even a chested effort, standard shot or anything else), the pass that led to it, the general build-up play, what players are around the shooter and between them and the goal, and the positioning of the goalkeeper.
Opta look at 20 variables in all, but one example that can help people understand xG better, and the simplest type of chance to assess is a penalty. Here, the stats can use the raw data and look at one simple question: how often are penalties scored on average?
The Opta model assigns penalties an xG of 0.79, with a figure of between 0.75 and 0.80 used by all main models. As said, nobody claims xG is perfect, and even this has issues – chiefly that the xG varies according to the time in the match and the status of the game (there is more pressure on a last-minute spot kick when the scores are level), as well as in shootouts versus normal time.
xG Variants and Related Stats
When we talk about xG, most of the time we are referring to the total xG of a team over the course of a game. For example, on the 25th of April 2026, Wolves lost 1-0 to Spurs and in a game of few chances, this was an accurate reflection of the xG.
The combined xG from all the shots that Wolves had was 0.70, while for Spurs it was 0.92. Tottenham didn’t quite reach an xG of one, but with 0.92, they would have been expected to score one most of the time. Wolves had a lower expected goals total of 0.70, so they were more likely than not to score, but on this occasion, they failed.
Expected goals can also refer to the chance of a given effort being a goal. So if a player produces a particularly brilliant finish, one might comment that the xG was just 0.05. There are also a whole host of similar or related stats.
- xG open play – xG created from open play
- xG set play – assesses chances from set pieces
- xGOT – expected goals on target is an assessment of the quality of the shot, rather than the chance. It is only relevant to shots on target
- xA – expected assists illustrates the chances of a pass leading to a goal and thus actually becoming an assist
- xG prevented – relevant to goalkeepers, this is the difference between the xGOT a stopper faces and the goals they actually concede
- Player xG – a player’s total xG in a game, their average per 90 minutes, or their total over a season