Expected goals (xG)
Glossary
As of 13 June 2026, expected goals (xG) is a football metric that estimates how many goals a team should have scored based on the quality of the chances it created, not the actual scoreline. Each shot is assigned a probability of scoring between 0 and 1; summed, they give a team's xG for the match. It is a descriptive statistic, not a forecast.
Updated · ScoreGPT
How is expected goals (xG) calculated?
xG is calculated shot by shot: every shot gets a scoring probability between 0 and 1, and those values are summed for the team's match xG.
- Every shot gets a value between 0 and 1 — the estimated probability that an average player scores from that situation.
- The value depends on the chance — distance and angle to goal, body part (foot vs head), whether it was a one-on-one, the type of assist, and more.
- Sum the shots to get a team's match xG.
A tap-in from the six-yard box might be worth 0.7 xG; a speculative 30-yard effort 0.03. If a team's shots add up to 2.3, its xG is 2.3 — regardless of whether it scored zero or four.
What does an xG of 2.3 mean?
An xG of 2.3 means the chances a team created were collectively worth about 2.3 goals against an average finisher — independent of how many it actually scored. The gap between goals scored and xG shows over- or under-performance:
| Scenario | Goals scored | xG | Interpretation |
|---|---|---|---|
| Underperformed | 1 | 2.3 | Bad finishing, good goalkeeping, or variance |
| Overperformed | 4 | 2.3 | Clinical finishing or favourable variance |
xG is most useful over many matches, where finishing luck evens out, rather than from a single game. That is why analysts treat a one-off xG gap with caution.
Why do AI models and analysts use xG?
xG separates process (chance creation) from outcome (goals), which can be noisy — so it can reveal when a result does not reflect the underlying performance. A team can lose 1–0 while generating 2.5 xG to 0.4, a sign the scoreline may flatter the opponent.
Definitional pages from analytics providers such as Opta and FBref have made xG mainstream, and AI prediction tools — including ScoreGPT's models — can weigh chance-quality signals like xG alongside form, injuries, and context when reasoning about a match. xG itself is a descriptive statistic, not a prediction.
What is the difference between xG and goals?
Goals are what actually went in; xG estimates how many goals the chances were worth — the table below contrasts them.
| Metric | What it measures | When it's known | Why it differs |
|---|---|---|---|
| Goals | The actual balls that crossed the line | Known at full time | The literal result |
| xG | Average goals the chances created were worth | Estimated post-match from shot quality | Driven by finishing, goalkeeping, and variance |
The gap between the two is what shows over- or under-performance over time.
Frequently asked
▸Is a higher xG always better?
Generally a higher xG means a team created better chances, which is a good sign. But xG describes chance quality, not the result — a team can have high xG and still lose. It is most reliable across many matches, not one.
▸What's the difference between xG and goals?
Goals are what actually went in. xG estimates how many goals the chances created were worth, on average. The gap between them shows over- or under-performance, often driven by finishing or goalkeeping.
▸Does ScoreGPT use xG in its predictions?
ScoreGPT's five AI models reason over many signals when analysing a match, which can include chance-quality metrics like xG alongside form, injuries, and context. xG is one input, not the whole method.
▸Is xG a prediction?
No. xG is a descriptive statistic about the quality of chances in matches already played. Prediction tools can use it as one input, but on its own it forecasts nothing.
Get ScoreGPT free
For information and entertainment only.