Expected Goals (xG)
Expected goals assigns a probability to every shot, learned from a historical archive of attempts and their outcomes, and sums those probabilities into a description of chance quality.
What it means
Each attempt is described by its features and valued by the frequency with which comparable attempts became goals. The sum over a match or a season describes how much chance quality a side produced.
Why it matters
It separates the quality of a chance from the fact of a goal, which makes performance readable over samples too small for goals to be informative: a match, a month, a player's first season.
Where it is used
Team and player performance review, recruitment screening and post-match analysis across every league covered by a data provider. It is the base layer on which the rest of the modern metric family is built.
Where it misleads
It is deliberately blind to who took the shot, so finishing skill is excluded by design. Over a single match it is dominated by variance, and without the game state it flatters sides that were chasing the game.
Key points
- The output is a probability, so summed values are expected goals and not goals.
- Distance and angle carry most of the weight; every other feature refines them.
- It is trained on an average player, so above-average finishing is invisible to it.
- Summed over a season it is stable; over one match it is noisy.
- Provider definitions differ, so totals from two pipelines are not interchangeable.
- Always read the total beside the shot count, never in isolation.
Expected goals describes the chance, not the finisher, and reading it as a verdict on a player is the most common way the metric is misused.