A shot estimate, not a guaranteed goal
Shot-based expected goals estimates the probability that a shot becomes a goal, given the information available to a particular model. Location and angle are common inputs. Some models also use the positions of defenders and the goalkeeper. Two providers can assign different values to the same chance because their features and training data differ.
Adding shot values gives an expected goal total. It does not describe the number of goals that must occur. A team can score from a low-probability attempt and fail to score from several good opportunities. That is not, by itself, proof that the model failed.
Worked example: an expectation is not a probability
Imagine three shots valued at 0.10, 0.20 and 0.30. Their total expected goals is 0.60. It would be wrong to say that this automatically means a 60 percent chance of scoring at least once. Under a simplifying independence assumption, that chance would be 1 minus 0.90 times 0.80 times 0.70, or 49.6 percent.
Real shots are not always independent. A rebound can occur only after the first attempt does not finish the attack. Game state can also change later opportunities. Use the calculation to understand the distinction between expectation and probability, not as a universal method for combining every shot sequence.
Use comparable evidence
Keep provider, model version, time period and competition coverage consistent when comparing teams. Penalty inclusion, own-goal handling and post-shot models can change the meaning of the total. A post-shot estimate can include the placement of a shot; a pre-shot estimate answers a different question.
Do not describe Match Insights as a shot-level xG provider unless a verified source supports that feature. A scoring rate calculated from match results is not interchangeable with shot-based xG. Missing shot data should remain missing. Substituting a different metric under the same label creates false precision.
Common question
Can a team score more goals than its xG?
Yes. Expected goals is an estimate of chance quality, not a limit on the actual score.
Sources and further reading
Continue your research
- How a Poisson football score model works
- How to compare football form without ignoring the opponent
- A football match research checklist