What Is xG (Expected Goals) in Football and How to Read It

What expected goals (xG) actually measure

Expected goals, usually written xG, is a statistic that estimates the quality of a shot. Every shot receives a value between 0 and 1. That value is the probability that a typical shot taken in the same situation ends up as a goal. A shot with an xG of 0.10 is one that, on average, scores about once in ten attempts. A shot with an xG of 0.80 scores about eight times in ten.

Add up the xG of all shots a team takes in a match and you get that team's match xG. It answers a simple question: given the chances created, how many goals would we normally expect? It does not say how many goals the team deserved in a moral sense. It describes the quality and quantity of chances, nothing more.

How an xG model is built

xG models are trained on very large collections of past shots. For each shot, the model looks at its characteristics and checks how often similar shots became goals. The most common inputs are:

• Distance from goal: the closer, the higher the value.
• Angle to goal: a central shot sees more of the goal than one from a tight angle.
• Body part: headers usually score less often than shots with the foot from the same spot.
• Type of assist or situation: through ball, cross, cutback, set piece, rebound, fast break.
• Penalty or open play: penalties are treated as their own category, and are commonly valued at roughly 0.75 to 0.80.

Some providers also use the position of defenders and the goalkeeper, when tracking data is available. Because each provider builds its own model, the same match can show slightly different xG values depending on the source. That is normal. What matters is to compare numbers from the same model.

A worked example, shot by shot

The values below are illustrative figures for an imaginary team, chosen to show how the calculation works. They are not data from a real match.

This team created chances worth about 1.5 goals. If it scored 0, 1, 2 or 3 goals, all of those outcomes are plausible. The total only tells you the middle of the range. Football is a low-scoring sport, so randomness has a large influence on single matches.

How to read the xG of one match

When you see a final score and an xG line side by side, compare them in three steps.

• Compare the two xG totals. If one team has 2.1 and the other 0.6, the first side clearly created better chances.
• Compare xG with the goals scored. A team that scored 3 from 1.0 xG was probably efficient or lucky. A team that scored 0 from 2.5 xG was probably wasteful or unlucky.
• Look at the shot list. One huge chance (a penalty for example) can make a total look better than the open-play performance really was. Ten shots of 0.05 give 0.5 xG, but that is a very different story from one single 0.50 chance.

A single match is a small sample. A 1-0 win with 0.4 xG against 1.8 xG is a result worth noting, but one such game says little about the long-term strength of either team.

Reading xG over a season

xG becomes more useful when you accumulate it. Two versions are especially common:

• xG for (xGF): the quality of chances a team creates.
• xG against (xGA): the quality of chances it concedes.

The difference between them (xGD, or xG difference) summarises overall performance. Over a long period, xG difference tends to track a team's underlying level better than the league table does, because it removes part of the luck from finishing and goalkeeping. If a team is far above its xG for several months, it is usually scoring more than its chances justify, and a drop is more likely than a rise. Comparing a team's actual goals with its xG gives a rough idea of overperformance or underperformance, although a few elite finishers can hold a small positive gap for years.

From xG to match probabilities

Match xG is not a score prediction, but expected-goal values for each side can be turned into probabilities. A classic method is the Poisson distribution, which describes how often a given number of goals occurs when the average is known.

Here is a simplified example, with made-up inputs: home team expected goals of 1.5, away team expected goals of 1.0. Running the Poisson calculation for both sides and combining the results gives approximately:

Note that even the favourite wins only about half the time, and a draw remains a real possibility. This is why a probability should be read as a degree of likelihood, not as a verdict. AI Football Predictor publishes probabilities and xG for every match, so you can see how these figures relate to each other on real fixtures.

Limits of xG you should know

xG is helpful, but it is a model and it has blind spots.

• It ignores finishing skill in the basic version. The model gives the same value to an average shooter and to an exceptional one in the same position. Over many shots, great finishers can beat their xG.
• It does not judge the shot placement. A related metric, post-shot xG, looks at where the ball was aimed and how fast it travelled, and is used to evaluate goalkeepers.
• Game state matters. A team leading 2-0 may sit back and allow low-quality shots, while a team trailing may take more speculative efforts. This can inflate or deflate totals.
• It does not know about injuries, tactics or motivation unless these show up in the shot data.
• Models differ. Do not compare numbers from two providers as if they were identical.

Common mistakes when using xG

• Treating xG as a forecast of the exact score.
• Drawing strong conclusions from one match.
• Ignoring that a penalty or a single big chance can dominate the total.
• Believing that a team with high xG must win. Chances can be missed, and the opponent can score from a single moment.

If you use statistics in connection with betting, remember that no statistic guarantees an outcome. Gambling is for adults only and carries real financial and personal risks. Treat xG as a tool to understand football, not as a promise of results.

The short version

xG converts shots into probabilities of scoring. Add them up to measure the chances created, compare them with the real score to see where luck or efficiency played a part, and prefer long samples to single games. Used with care, it gives a clearer picture of how a team really performs than the scoreline alone.