Expected goals, or xG, is the most useful number in football and the most abused. Used well, it tells you whether a team is creating and conceding good chances. Used badly, it becomes a stick to beat managers with after a single match. The gap between those two uses is worth understanding.
What the number actually is
xG assigns every shot a probability of becoming a goal, based on the historical outcomes of thousands of similar attempts. A chance worth 0.1 xG is one that gets scored roughly once in ten tries. Distance, angle, body part, the type of pass that created it, whether it was a clear sight of goal or a hopeful poke, all of it feeds the estimate.
So a scoreline of 1-0 might hide an xG story of 2.4 to 0.6. That doesn't mean the wrong team won. It means the team that lost created the better chances, on this occasion, and finishing variance did the rest.
xG doesn't tell you what happened. It tells you what usually happens from the chances that were created.
Where people go wrong
Three mistakes recur:
- Reading it over one game. A single match is far too small a sample. Finishing swings wildly from game to game; the signal only emerges over ten, twenty, thirty matches.
- Confusing xG with deserving to win. Football rewards goals, not chance quality. xG is a description of process, not a verdict on justice.
- Ignoring who is shooting. The model assumes a league-average finisher. An elite striker will, over time, out-perform his xG, and that outperformance is a real, repeatable skill, not luck.
The right questions to ask
Used properly, xG is a diagnostic. Over a decent run of games:
- Is a team's xG for far above its actual goals? It may be creating well but wasting chances, often a fixable finishing or personnel issue.
- Is its xG against creeping up while results stay good? That's a warning the goalkeeper or variance is papering over defensive cracks.
- Are the chances themselves good, or is a team inflating its xG with lots of low-value long shots? Volume without quality flatters the total.
Beyond the shot
The frontier has already moved past shot-based xG. Newer models measure the value a team adds by moving the ball into dangerous zones even when no shot results, possession-value and expected-threat models that reward the pass into the half-space as much as the shot at the end of it. They exist because xG has a blind spot: a team can dominate territory, strangle a game, and take few shots by design, and pure xG will undersell it.
The lesson isn't that xG is wrong. It's that no single number closes a football argument. xG is a lens, a very good one, and like any lens it shows you more when you know what you're looking at, and misleads you when you point it at the wrong thing.