Expected goals (xG)

Expected goals (xG) assigns each shot in a match a value between 0 and 1: the historical probability that a shot with similar characteristics resulted in a goal. Inputs typically include distance and angle from goal, whether the shot was taken with foot or head, the type of pass that created it, and defensive pressure at the moment of the shot. Summed across a match or season, xG measures the quality of chances created and conceded, separate from whether they were actually scored.

Models are usually trained with logistic regression or gradient-boosted trees on large shot samples drawn from player tracking data and event feeds. This differs from a raw shot count, which treats a tap-in and a speculative long-range strike as equally dangerous. Unlike a whole-team rating system such as elo rating, xG operates at the level of the individual chance. Because providers train on different shot samples and features, xG for the same match often differs between sources.

Analysts use xG to separate finishing skill and luck from underlying chance quality, informing recruitment, tactics and commentary in soccer and, in adapted form, hockey; it grew out of the same empirical tradition as baseball's sabermetrics. The common misreading treats xG as a precise "deserved" score rather than a probabilistic estimate: a team can beat or miss its xG for an entire season through finishing skill, variance, or factors the model does not capture, such as goalkeeper positioning.