What Is Expected Goals (xG) in Soccer?

Not all shots are equal. A tap-in from two yards is far more likely to be scored than a shot from 30 yards. Expected goals quantifies that difference by looking at thousands of past shots with similar characteristics and measuring how often they became goals. A shot with an xG of 0.3 is one that, historically, was scored about 30 percent of the time.

Common inputs include shot distance and angle to goal, whether the shot was a header or with the foot, the type of assist, such as a through ball or cross, whether it came from open play, a set piece, or a counterattack, and in more advanced models, defender and goalkeeper positions. Penalties are usually assigned a fixed value based on historical conversion rates.

Adding up the xG of all a team's shots in a match gives its expected goals. Comparing team xG with xG against shows which side created better chances. A team that wins 1 to 0 while trailing 0.6 to 2.1 in xG was probably fortunate; over many matches, xG difference tends to predict future results better than actual goal difference.

For players, xG shows how many goals their chances would typically produce. A striker who scores many more goals than their xG may be an exceptional finisher, or may be on a hot streak that will regress. Over large samples, some players consistently outperform xG, but the gap is usually smaller than short-term numbers suggest.

Expected assists, xA, measures the xG of shots created by a player's passes. Non-penalty xG, npxG, excludes penalties so that players are not credited for being their team's penalty taker. Expected goals on target, xGOT, measures shot quality after the shot is taken, using where it was placed in the goal.

Goals are rare and random in soccer, so match results are noisy. xG captures how well a team created and prevented chances, which is more stable over time. Analysts, clubs, and betting markets use xG to assess performance, identify regression candidates, and inform predictions.

xG models differ between providers, so numbers are not always comparable across sources. Basic models ignore defensive pressure and goalkeeper position. xG does not credit chances that never became shots, such as a dangerous cross cut out at the last moment. Small samples, like a single match, can mislead.

Look at xG over several matches rather than one. Compare xG for and against, and check non-penalty figures. Use differences between goals and xG as a prompt for questions, not conclusions: is a team finishing unusually well, or is something the model misses, like set-piece design, driving results?

Markets already use xG and similar metrics heavily. A team outperforming its xG may be overrated by casual bettors, but prices from sharp books usually account for underlying numbers. xG is most useful in a model that also considers squad changes, schedule, and market prices.

Across a full season, team xG difference is one of the strongest indicators of underlying quality. Teams whose points total is far above what their xG suggests often fall back the following season, while teams underperforming their xG often improve. This is regression toward the mean, and it is one of the most reliable patterns in football analytics.

xG against measures the quality of chances a defence concedes. Comparing goals conceded with xGOT faced, rather than xG, isolates goalkeeper performance, because xGOT accounts for where shots on target were placed. A goalkeeper conceding fewer goals than xGOT faced is saving more than an average keeper would.

Penalties have high, fairly stable conversion rates, so many analysts use non-penalty xG to compare open-play performance. Set pieces, such as corners and free kicks, produce many low-xG headers; teams with strong set-piece routines can create more value than basic models capture.

Start by comparing a few teams' xG for and against over a season, then look at players' non-penalty xG per 90 minutes. Use one data source throughout, and compare across at least ten matches before drawing any firm conclusions.

This page describes what the metric measures and is not betting advice.