AIGOALDATA / Guide

Understanding football xG

Expected goals estimate chance quality under a model. They describe shot context, not a guaranteed final score.

What are xG?

xG, short for “expected goals”, aim to measure the quality of chances rather than the number of goals alone. The idea is to give each shot a value between 0 and 1: the probability that a shot of that type becomes a goal, based on a very large number of comparable shots observed in the past. The factors taken into account vary between data providers, but they often include distance, angle, the body part used, the type of assist or the phase of play.

A close-range shot at an empty goal is worth a lot; a long-range effort from a wide angle is worth very little. A penalty, which is converted roughly three times out of four, has a high value. Adding up the values of all a team’s shots gives its xG for the match: an estimate of how many goals an “average” team would have scored with the same chances.

An example to understand

Imagine a purely fictional match that ended 0-1. The losing team shot sixteen times for a total of 2.1 xG, including several clear chances. The winning team shot only four times, for 0.5 xG, and scored from one of its rare chances. The score says the second team won; the xG say the first team created clearly better chances.

Neither number is “wrong”. The score tells what happened, and it is what counts in the table as in every betting market. The xG describe the quality of the play produced. Over a single match, the gap can be huge. Over a whole season, the two measures tend to converge, which makes xG useful for judging how solid a team really is.

What xG measure well

Over a long period, xG are generally more stable than goals. A team can go through a run of matches where it converts everything, or misses everything; such runs often end up correcting themselves. Comparing goals scored with xG produced therefore helps to spot a team whose results exceed or understate its real level of chance creation.

xG conceded, sometimes called xGA, apply the same logic to defence: they measure the quality of the chances allowed to the opponent. A team that concedes few goals while allowing a lot of xG may owe a great deal to its goalkeeper, or to a little luck.

Their limits

xG are not raw data: they come from a model, and each provider uses its own. Values can therefore differ from one source to another for the same match. They only count shots: a dangerous cross that nobody meets, or a counter-attack wasted by a poor final pass, does not appear.

They also depend on the game state. A team that is ahead often sits deeper and concedes long-range shots, which can inflate the opponent’s shots without real danger; a team that is behind multiplies desperate attempts. Finally, they only imperfectly account for individual quality: an elite striker converts more often than an average player in the same situation, and a great goalkeeper stops shots another would let in.

It is also worth remembering that a match’s xG do not account for what happens after a missed shot: a team that wastes its chances early on may change its approach, and the game may follow a completely different course.

Indicators related to xG

xG have given rise to a whole family of indicators. xA, or expected assists, credit the passer with the value of the chance they created, whether or not the shot is converted. Some providers also offer a measure of the quality of shots on target, which takes into account where the ball is heading in the goal: it helps to judge goalkeepers’ work.

Simpler measures remain useful: the number of shots, the share of shots on target or the number of big chances. None is perfect. Each sheds light on part of the game, and it is reading them together, over a sufficient period, that gives a faithful picture of a team.

Reading a match with xG

After a game, xG help to put a score into perspective. A team beaten despite a clear xG advantage often lacked luck or precision; a team that won with very little xG may have benefited from a single incident. In an isolated match, such gaps are frequent and nothing out of the ordinary.

The value shows mainly through repetition. If a team wins match after match while creating fewer chances than its opponents, its results are likely to deteriorate. If another regularly loses while dominating on xG, its results have a good chance of improving. xG then serve as a trend indicator, not a verdict.

xG and predictions: keeping them apart

xG measured after a match describe the past. A prediction concerns a match that has not yet been played, against a specific opponent, at a specific venue, with a squad that may have changed. Past xG can inform your thinking, but they do not transfer as they are to the next game.

On AIGoalData, the “expected goals” shown for an upcoming match are an estimate calculated before kickoff, for the whole game and for each team. They do not come from counting shots, since those shots do not exist yet. They help to picture the expected scenario, just like the probabilities; they are never presented as a final score.

Good practice

Compare xG with actual goals over ten matches or more rather than over a single game. Always check the source, the period and the competition before comparing two figures: xG from different models cannot be compared directly. Look at xG produced and conceded together, because a team is judged on both ends of the pitch.

Finally, keep in mind that an isolated xG figure does not, on its own, justify BTTS, a match with over 2.5 goals or a straight winner. It is one indicator among others, useful for understanding a match, never a guarantee of a result. Betting remains risky and is forbidden to minors; our Responsible gambling page sets out sensible habits.

Predictions Home