The final whistle says 1–0. The statistics panel says the losing team produced more expected goals. Both statements can be true, because they describe different things: the score records what happened, while xG estimates the quality of the shots that were taken.
The number becomes useful when it sends you back to the match with a better question. It becomes misleading when it is treated as proof that the “wrong” team won.
Start with one shot
The Premier League’s introduction to expected goals explains xG as a probability assigned to a shot, on a scale from zero to one. A higher value describes a chance that the model considers more likely to become a goal. It does not promise that this particular player will score.
For an illustrative example, imagine one chance rated 0.5 and five chances rated 0.1 each. Both groups total 0.5 xG, but the attacking sequences looked very different. The totals alone do not show you whether the team created one clear opening or repeatedly shot from difficult positions.
A total is not a probability of scoring exactly that many
A team finishing with 1.7 xG has not been awarded 1.7 theoretical goals. Nor does the total by itself tell you its probability of scoring at least once. Adding shot values is a summary of the opportunities assessed by the model, not a replacement result.
Models can use different information and produce different estimates. When comparing matches, use a consistent provider and check what its measure includes. Do not mix ordinary xG, post-shot models and penalty-adjusted figures as though their labels were interchangeable.
Put the chances back into the match
Ask when the shots occurred. Did one side create openings while the score was level, or spend the final twenty minutes attacking a team protecting a lead? Was there a sending-off? Did the apparent volume come from a short sequence of rebounds?
These observations do not cancel the statistic. They explain the football around it. Watch a few relevant passages again and note the pass, movement and defensive situation that produced the shot. That is more informative than repeating the final xG total with stronger language.
Use several matches to test an impression
One match contains plenty of variation. If you think a team consistently creates good chances but finishes poorly, examine a longer run using the same data source, then look at the actual opportunities. Injuries, opponents and changes in role may also affect the pattern.
The official Premier League statistics explanation provides context for its data. Keep the score, the model and your viewing notes separate. A useful conclusion sounds like a football observation supported by evidence, not a claim that a decimal number has rewritten the result.