Expected goals is a season metric being used as a match metric

Expected goals assigns each shot a probability of scoring based on location, type and situation. It is the most useful public football metric and it is routinely misapplied.

What it is good at

Over twenty or more matches it describes whether a team is generating and conceding good chances, and it predicts future goal output better than past goal output does.

That is a real finding and it is why every serious recruitment department uses it.

What it is bad at

A single match. One game contains perhaps 25 shots, and a metric built on probabilities needs far more events to be meaningful. A team can lose 2-0 having “won” the xG and the number tells you very little.

Quoting a single match’s xG as evidence that a result was unjust is the most common misuse in football coverage.

What the model cannot see

Pre-shot passing beyond basic flags, so a shot after a cross-goal pass is priced like the same shot after a cycle. Defensive pressure at the moment of the shot. The goalkeeper’s position. And any move that created a chance somebody fumbled before shooting.

These wash out across a season. They do not wash out in one game.

Using it properly

Treat xG as a description over a stretch and a weak one over a match. If someone quotes a single game’s figure as proof of anything, they are using a season tool on a sample it was never built for.

Reported at ESPN; analysis ours.