Expected goals: what the metric still misses

Expected goals entered everyday football talk because they answered a fair question: how good were the chances, not only how many finished in the net? Used carefully, the metric separates clinical finishing from a side that rarely reaches dangerous zones. Used carelessly, it becomes a verdict on a coach after three matches.

Models weight location, assist type and sometimes defensive pressure. They do not fully capture a striker's weaker foot under fatigue, a goalkeeper's starting position on a specific set piece, or the strategic choice to protect a one-goal lead by refusing low-percentage shots. Those omissions matter when narratives harden too quickly.

Game state and shot selection

A team chasing a deficit takes different shots from a team managing a lead. Aggregate xG can look similar while the underlying decisions differ. Analysts who ignore game state risk praising volume that was born of desperation. Coaches already know this; the public conversation often does not.

Set pieces create another blind spot. A well-rehearsed routine can produce a high-value chance that looks ordinary in a generic model if the marking scheme is not encoded. Conversely, a scramble rebound may inflate a figure that owed more to luck than design. Film still sits beside the spreadsheet for a reason.

The healthy use of expected goals is comparative and patient. Look at trends over a block of matches, pair them with chance creation maps, and ask whether the process matches the coach's stated plan. When the metric and the eye disagree, the answer is rarely to discard one tool. It is to ask a better question about what the model cannot see.