What is xG in football? Expected goals explained

Think of many comparable attempts, not a countdown to the next goal. A model assigning 0.20 xG estimates roughly 20 goals per 100 such shots on average. Any individual attempt still produces either a goal or no goal. It cannot produce a fifth of a goal.

Four misses do not make the fifth shot certain to go in. Nor does a missed chance lose its pre-shot xG afterwards. The number describes the opportunity; the outcome tells you what happened this time.

Providers train statistical models on recorded shots. Common inputs include distance, shooting angle, body part and the pass or action before the attempt. Some models also capture the goalkeeper, defenders and pressure around the shooter. The model returns a probability between zero and one for that shot.

There is no universal distance-to-xG chart: a crowded header and an unopposed shot from the same spot are different chances. Different providers and model versions can give different values to the same attempt. Compare a team's numbers within one provider and the same definition. The examples here use assigned teaching values, not a Match XI shot-rating model.

Eight attempts worth 0.10 each add up to 0.80 xG. One attempt worth 0.80 also gives 0.80 xG. The totals match, but the opportunities are not interchangeable. In this deliberately simplified example, the single big chance has an 80% chance of producing a goal; the eight smaller, independent chances have about a 57% chance of producing at least one.

Why? The eight-shot set can produce several goals, whereas the single shot can produce at most one. Both sets have the same expected number of goals, but different possible outcomes. Summing xG is not the same calculation as finding the chance of scoring at least once.

Consider a fictional match: Blue lose 0-1 despite recording 1.80 xG to Gold's 0.60. You know the actual result and which side accumulated the larger xG total. You do not know that Blue deserved a 2-1 win, that their striker was solely to blame, or that the next match will correct the result.

A total hides the individual chances and their timing. Finishing, saves and ordinary variation can separate goals from xG. The score also changes how teams play: a side protecting a lead may allow more later attempts. Look at the attempts and match events before turning the total into a verdict.

The league table rewards the actual result. An xG advantage does not add points, alter goal difference or break a tie in the standings.

Ordinary xG assesses the situation before the shot's outcome is known. Post-shot models add information about the attempt itself. Opta's expected goals on target, xGOT, evaluates on-target shots using information including where the ball is heading in the goalmouth. Other providers use names such as post-shot xG or PSxG, with their own model definitions.

A difficult chance struck towards a corner can therefore have a low pre-shot xG but a much higher post-shot value. An excellent chance blazed wide still contributes to ordinary xG, but it is not an on-target attempt for xGOT. Neither number is simply a count of shots on target.

For goalkeeper shot-stopping, post-shot information helps distinguish a routine attempt from a difficult save. Do not judge a keeper from the opponent's pre-shot xG alone, or assume one post-shot number measures every part of goalkeeping.

Penalties normally contribute to total xG, but not to non-penalty xG, often labelled npxG. The assigned penalty value depends on the model: the cited Opta explainer uses 0.79, while Hudl Statsbomb's explanation gives 0.78 for its 2022 update. Neither is a law of football or a guarantee that a particular player will score.

Ordinary shot-based xG is not limited to shots on target. An attempt blocked or sent wide can still receive a pre-shot value. But a dangerous pass that nobody turns into a recorded shot creates no shot xG of its own. That is one reason xG does not describe every threatening attack.

Rebounds need context too. If a second attempt exists only because the first was saved, those are not independent opportunities available regardless of the first result. Do not feed such a sequence into the independent-chance calculation above and call the result a match forecast.

Start with the label: team or player, one match or a season, total or per 90 minutes, penalties included or excluded? Then check the provider. A 0.60 per-90 figure and 0.60 across several substitute appearances are not the same claim.

For a fictional player with 3 goals from 1.50 xG, goals minus xG is +1.50. That describes the recorded period. It does not prove sustainable finishing skill or mean the next chances must be missed. For a player below xG, the missing goals are not a debt that must be repaid.

Finally, connect the stats to participation. A productive recent spell does not establish that a player will start the next match. Use confirmed team sheets for actual starters, clearly labelled predictions before confirmation, and match records for what happened. Match XI's teaching examples above are not live xG statistics for the clubs linked below.