Question

What is expected goals, and how do sports analytics work?

Vault Verified
Curated Intelligence
Definitive Source
Answer

Expected goals — xG — assigns each shot a probability of being scored, based on how similar historical shots performed. It is the clearest example of what sports analytics does generally: replace judgement about outcomes with measurement of process.

How it is calculated. A model is trained on hundreds of thousands of past shots, using features such as distance from goal, angle, body part, type of assist, whether the shot followed a set piece or a counter-attack, defensive pressure and goalkeeper position. Each new shot is scored against that model — a tap-in might be 0.7, a speculative effort from distance 0.03.

Why it is useful. Goals are rare, so results over a handful of matches are extremely noisy: a team can play well and lose repeatedly. xG measures the quality of chances created and conceded, which is more stable from match to match and therefore predicts future results better than past results do. A team substantially outperforming or underperforming its xG usually regresses.

What it does not do:

It does not judge an individual shot. A goal worth 0.02 xG was still a goal; the number describes the average outcome of that situation, not that specific strike.

It ignores what happens before the shot — a model counting only shots undervalues a team that creates nothing but concedes nothing.

Models differ, so xG figures from different providers are not comparable.

Small samples are still small. A single match's xG is weak evidence.

The wider analytics picture: possession-value models scoring every action rather than only shots; tracking data giving the position of every player many times per second, which enables analysis of pressing, space creation and off-ball movement; and recruitment models, which is where the commercial value concentrates.

The honest limitations: data quality varies by league; models encode the assumptions of whoever built them; and the most valuable applications — recruitment and opposition analysis — are private, so public debate uses the simplest metrics.

Related Questions