The Problem with xG: Why Expected Goals is a Flawed Metric

Expected Goals (xG) are everywhere in football analysis, presented as the primary measure of performance. But really? This increasing reliance on single numbers hides significant problems: the problem with xG is that it oversimplifies the beautiful game, creating a distorted view of what is actually happening on the pitch. Although intended to measure shot quality, Expected Goals has several fundamental flaws that limit its use and can lead to inaccurate conclusions about players and teams.

Here are the key xG issues that every fan, analyst and coach should understand.

1. Statistical Problems: Small Samples and High Variance

The first big problem with xG lies in its statistical foundation. Football is a low-scoring game where a typical shot only has about a 10% chance of being a goal. This creates huge statistical noise. For a player’s performance to differ statistically from random chance, they need to take a large number of shots. Since most players do not achieve this volume, the gap between their actual goals and their xG is often meaningless. This high variance makes it nearly impossible to use xG to reliably assess a player’s finishing skills over a short period of time.

2. The “Shooter’s Blindness” Problem: xG Ignores Player Skill

Perhaps the most glaring problem with xG is its “shooter blindness”. The metric assigns the same value to a shot, regardless of who takes it. A clear chance has the same xG whether it falls to a world-class striker or a centre-back. This completely ignores the realities of football, where finishing abilities vary dramatically from one player to another. Clubs pay millions of dollars for elite strikers precisely because they are better at converting chances than “average” players. By treating every shooter the same, xG fails to measure the skills it is often used to evaluate.

3. Context Matters: xG Exists in a Vacuum

Football matches are dynamic and complex, but xG analyzes moments in isolation. This creates a significant context problem.

• This Ignores Game Buildup: xG only measures the shot itself, not brilliant dribbles or perfect passes that create opportunities. Dangerous attacks thwarted with a final tackle do not receive xG credit.

• This Misses Game Status: This metric does not take into account psychological factors such as pressure, momentum, or fatigue. Penalties in the final minutes of a final match are treated the same as penalties in the opening moments of a friendly match.

• This Ignores the Goalkeeper: Most xG models do not take into account the quality or position of the goalkeeper. A high xG shot saved by a world-class goalkeeper is considered a miss, thereby imposing an unfair penalty on the shooter’s performance.

4. Team Quality Issues: Skewed Results for Top Teams

When you analyze xG over a full season, another problem emerges. Top teams with elite strikers consistently score more goals than their xG predicts. Conversely, struggling teams often score less. This happens because the xG model is based on an “average” finisher, but top teams do not have average finishers. This systematic bias means xG can be a poor tool for comparing teams with different levels of quality, as xG consistently underestimates the efficiency of the best teams and overestimates the potential of the worst teams.

5. Single Game Problem: Misusing xG for Match Analysis

One of the most common misuses of metrics is their application to single games. The problem with xG is that it is a tool for analyzing long-term trends, not for assessing 90 minutes of matches. Randomness plays a big role in any one game. Stating that a team “deserves to win” because they have a higher xG is a misapplication of probabilistic metrics. A high XG in a loss doesn’t necessarily mean a team is unlucky; it’s just a data point in a very varied sport.

While Expected Goals can be a useful part of a broader analytical toolkit, these problems demonstrate why Expected Goals should not be a determinant of performance. The true story of a game is told in its context, skill and human drama—elements that can never be fully captured by a single number.