Expected goals against is a team stat, not a keeper stat
Expected goals against has become one of the most cited defensive numbers in football, and for good reason: it grades the quality of the chances a side concedes rather than the fluky final score. But the way it gets used has drifted into a mistake. A goalkeeper with a low xGA behind him gets praised as if the number were his report card, and a keeper on a leaky team gets blamed for an xGA that measures his defenders’ failings, not his own. Expected goals against is a team stat. It is built entirely from the chances a team allows before the goalkeeper does anything, and it contains, by construction, zero information about how well he actually stops shots. Reading it as a keeper metric credits and blames the wrong man.
What xGA is built from
Expected goals against sums the expected-goal value of every shot a team concedes — each chance rated by its location, angle, the type of pass that created it, and the situation, then added up. Crucially, that value is assigned at the moment the shot is taken, based on how likely an average finish is from that spot. It is computed the same way whether the shot flies into the top corner or straight at the keeper, because it describes the chance, not the outcome. That means xGA is a measure of chance prevention — of how well eleven players defend — and the goalkeeper’s contribution to preventing the chance is only his share of the collective defending, not his shot-stopping at all.
The defense authors the number
A team’s xGA is written by its structure, its pressing, its positioning, and the caliber of attacks it faces. A side that defends deep and compact, funnels shots to low-value areas, and blocks the dangerous passes will hold opponents to a miserly xGA — and its keeper inherits that number regardless of whether he is world-class or ordinary. A side that bleeds high-quality chances through the middle hands its keeper an ugly xGA no matter how many miracles he pulls off. Two keepers of wildly different ability would post the same xGA behind the same defense, which is the clearest proof that the stat is measuring the ten players in front of the goal rather than the one in it.
Shot-stopping is exactly what xGA leaves out
The whole point of a goalkeeper is what happens after the shot is struck, and that is the one thing xGA does not see. Whether the keeper dives to tip a curling effort around the post or lets a stoppable shot squirm through his hands, the expected-goals value of that chance was fixed the instant the ball left the boot. xGA cannot reward a save or punish a howler, because both occur after its measurement is complete. A keeper could concede far fewer goals than his xGA suggests, through brilliant shot-stopping, or far more, through weak hands — and the xGA figure would be identical in both cases. The number is silent on the single skill a goalkeeper is paid for.
The comparison that actually grades the keeper
This is why the meaningful goalkeeping number is not xGA but the gap between xGA and goals actually conceded. If a keeper’s team allowed chances worth, say, forty expected goals and he conceded only thirty-two, that eight-goal difference is the shot-stopping contribution — goals prevented above what an average keeper would have allowed from the same shots. That post-shot comparison, often built from a post-shot expected-goals model that accounts for where each shot was placed, is the real keeper metric. Raw xGA is merely the baseline it is measured against. Quoting xGA alone as a goalkeeping grade is like quoting the difficulty of an exam without ever mentioning the score.
Variance and sample size cut both ways
Even the proper keeper number needs patience. Shot-stopping performance swings on small samples, and a keeper can run hot or cold relative to expectation for a stretch of matches on variance alone. A goalkeeper who saves several goals above expected over ten games has not necessarily found a new level, and one who leaks a few may not have collapsed. Over a full season the signal firms up, but early-season keeper numbers — and the xGA behind them — are noisy enough that confident verdicts are premature. The team stat and the keeper stat both stabilize slowly, and mistaking a hot run for a true rating is its own error on top of confusing the two stats in the first place.
Model choices sit underneath the number
xGA is not a single fixed truth but the output of an expected-goals model, and different providers build those models differently. What inputs a model weighs — shot location, assist type, defensive pressure, whether the effort was a header or a volley — varies from one system to the next, so the same passage of defending can carry a different xGA depending on whose model is crunching it. Set pieces, deflections, and rebounds are handled inconsistently across systems, and those are exactly the situations where chance quality is hardest to pin down. None of this makes xGA useless, but it means a team’s xGA carries a margin of error the tidy decimal hides, and small gaps between two sides’ figures may say more about the model than the defending.
What to read instead
To judge the keeper, read the shot-stopping gap: goals prevented versus post-shot expected goals, which isolates what he does with the shots that reach him. To judge the defense, read xGA and the chance-quality trends underneath it — how many high-value looks the side concedes and from where. Keep the two apart, and each becomes informative; fuse them, and you will keep praising keepers for their defenders’ work. StatLine’s Premier League tables and standings carry the team defensive context and the goals-conceded lines that let you separate the chance prevention from the shot-stopping, so an xGA figure sits where it belongs: as a verdict on the team in front of the goalkeeper.
The honest read
Expected goals against measures a real and useful thing — the quality of the chances a team concedes — and as a defensive team stat it is a clear upgrade on goals-against, which is hostage to finishing luck and goalkeeping alike. That is precisely why it belongs to the team and not the keeper: it is finished measuring before the goalkeeper ever touches the ball. A low xGA tells you a side defends well; it tells you nothing about whether the man in goal is elite or getting bailed out. Read xGA as a grade on the defense, and read the xGA-minus-goals gap as the grade on the keeper. Just never read the team’s expected goals against as the goalkeeper’s number, because his job starts exactly where the stat stops.