Expected Goals (xG) is a statistical metric that measures the quality of goalscoring chances by calculating the probability of a shot resulting in a goal based on historical data. Every shot taken in football is assigned an xG value between 0 and 1, where 0 means the shot has virtually no chance of being scored and 1 means it is essentially certain to result in a goal. The sum of all xG values for a team’s shots in a match gives their total match xG, representing the number of goals a team would be expected to score based on the quality of opportunities they created. Since its widespread adoption in the mid-2010s, xG has become the single most important advanced metric in football analytics.
How xG Is Calculated
xG models use machine learning algorithms trained on hundreds of thousands of historical shots to estimate the probability of any given shot resulting in a goal. The primary factors that influence a shot’s xG value include the distance from goal (shots closer to goal have higher xG), the angle relative to the goal (central shots have higher xG than shots from wide positions), the body part used (headed shots have lower xG than foot shots from similar positions), the type of assist (through balls and crosses create different types of chances), and whether the shot follows a fast break or a set piece.
A penalty kick has an xG of approximately 0.76, reflecting the historical conversion rate from the spot. A one-on-one with the goalkeeper from inside the six-yard box might have an xG of 0.60 to 0.80 depending on the specific angle and distance. A shot from the edge of the penalty area in a central position typically carries an xG of 0.05 to 0.10. A long-range shot from 30 yards might have an xG of only 0.02 to 0.04. These values represent the average probability across all players and situations — individual players may consistently outperform or underperform their xG based on their finishing ability.
Different data providers — including Opta, StatsBomb, FBref, and Understat — use slightly different xG models with different input variables and training data, which means xG values for the same shot can vary between providers. While these differences are generally small, they can affect team and player-level xG totals over a season. For most analytical purposes, the choice of provider matters less than consistent use of the same source for comparison, ensuring that like is being compared with like.
The concept of xG has expanded into several related metrics. Non-penalty expected goals (npxG) excludes penalties from the calculation, providing a cleaner measure of open-play and set-piece chance creation. Post-shot expected goals (PSxG) incorporates the placement and power of the actual shot, measuring not just the chance quality but also the finishing quality. Expected goals on target (xGOT) measures the quality of shots that hit the target, providing a measure of goalkeeper performance by comparing the goals conceded against the xGOT faced.
What xG Tells Us About Team Performance
Team-level xG analysis reveals performance information that raw results and goal tallies can obscure. A team that creates 2.5 xG per match but only scores 1.5 goals is likely experiencing bad finishing luck or facing exceptional goalkeeping, and their results are likely to improve as actual goals regress towards expected goals over time. Conversely, a team scoring 2.0 goals per match from only 1.2 xG is outperforming their chance creation through exceptional finishing or goalkeeping errors by opponents, and their scoring rate may decline as this luck corrects.
The difference between a team’s xG and their actual goals over a season is a powerful predictive indicator. Teams that significantly outperform their xG in the first half of a season tend to see their results decline in the second half as their scoring rate normalizes. Teams that underperform their xG in the first half tend to improve. This regression to the mean principle is one of the most reliable patterns in football analytics and provides a basis for predicting future performance that goes beyond simply extrapolating recent results.
xG difference (xGD) — the difference between a team’s xG created and xG conceded — is a better predictor of future league position than actual goal difference. Research has consistently shown that xGD explains more variance in future results than historical points, goals scored, or goals conceded individually. This makes xG an invaluable tool for prediction models, including those used for correct score forecasting, as it captures the underlying quality of a team’s performance rather than the noise introduced by finishing variance and random fluctuations.
xG for Individual Players
At the individual level, xG helps evaluate players’ goalscoring efficiency and chance quality. A striker who scores 15 goals from 12 xG is outperforming their expected output, suggesting elite finishing ability. A striker who scores 8 goals from 14 xG is underperforming, possibly due to a poor run of form, and may be expected to improve. However, sustained overperformance of xG across multiple seasons does indicate genuinely elite finishing — players like Lionel Messi and Robert Lewandowski have consistently outperformed their xG over their careers, demonstrating that finishing ability is a real and persistent skill.
The xG per shot metric reveals which players create or receive the highest-quality chances. A player with high xG per shot gets into dangerous positions close to goal, while a player with low xG per shot takes shots from less promising positions. This distinction is important for evaluating whether a player’s low scoring rate is due to poor finishing or poor chance quality. If the chances are low quality (low xG per shot), the issue is tactical or positional; if the chances are high quality but not being converted, the issue is finishing.
xG and Correct Score Predictions
Expected goals are the primary input for most correct score prediction models, including those used at Correct Score Predict. By estimating each team’s expected goals for a specific match, a Poisson distribution or similar statistical model can generate probability distributions for every possible scoreline. The higher the accuracy of the xG estimates, the more accurate the resulting correct score predictions will be.
xG data improves correct score predictions by providing a more stable and predictive measure of team quality than raw goals alone. A team that scored four goals last week from 1.2 xG had an exceptional day of finishing that is unlikely to be replicated; predicting them to score four again would be unwise. A team that scored one goal from 3.5 xG was unlucky and is likely to score more in future matches with similar chance quality. xG allows prediction models to see through the noise of individual match results and identify the underlying attacking and defensive capabilities that determine the most likely scorelines.
At Correct Score Predict, xG is central to our prediction methodology. We use expected goals data to estimate each team’s true attacking and defensive quality, adjust for the specific matchup and contextual factors, and generate the most accurate scoreline probability distributions possible. Understanding xG helps our users appreciate the analytical foundation behind our predictions and make better-informed betting decisions.
