Expected assists is a shot-quality stat, not a passing stat
Expected assists arrived to fix a real problem. The raw assist is hostage to finishing: a perfect through ball squandered by a striker earns nothing, while a scuffed pass turned into a wonder goal earns a full assist. Expected assists, or xA, cuts that cord by crediting the passer for the quality of the chance he created rather than for whether it was scored — it assigns each key pass the expected-goals value of the shot it set up. That is a genuine improvement, and xA deserves its place near the top of any creativity ranking. But the fix quietly relocates the measurement. xA does not grade the pass. It grades the shot the pass produced. And once you see that, a lot of what the leaderboard says starts to look like a shot-quality table wearing a passer’s name.
What xA actually measures
The mechanics are simple and worth stating plainly. When a pass leads directly to a shot, xA takes that shot’s expected-goals value—driven by its location, angle, and type—and hands the whole figure to the passer. So a creator’s xA is a sum of the xG of the shots his teammates took after receiving his passes. The number is built entirely from the properties of the shots, not from the properties of the passes. A pass’s difficulty, its disguise, the defenders it eliminated, the pressure the passer was under—none of that enters the calculation except insofar as it happened to produce a shot from a good spot. xA credits the passer for where his teammate ended up shooting from, which is only partly his doing.
Proximity to goal is doing the scoring
Because xA inherits the shot’s xG, it inherits xG’s single strongest input: distance to goal. A simple square ball rolled to a striker four yards out, who merely has to redirect it, generates a large xA because the resulting shot is high value. A breathtaking pass that slices through a back line but leaves the runner with a tight angle from eighteen yards generates a modest xA, because the shot it created was hard. The metric therefore rewards proximity—passes that end near the six-yard box—far more than it rewards the passing skill that manufactured the opening. The most valuable pass is frequently the one that turns a low-xG situation into a slightly-less-low one, and xA can barely see that pass at all.
Volume of service hides inside the total
Season xA is a counting stat before it is a rate, so it accrues to whoever pumps the most balls into dangerous areas. A wide player on a possession-heavy team who whips twelve crosses a game into the box will bank xA on the sheer volume of low-probability service, while a deeper creator who plays two perfect passes a match may trail him despite superior craft per attempt. This is not wrong, exactly—volume of good service is genuinely valuable—but it means the xA leaderboard blends two very different questions: who creates the best chances, and who creates the most. Reading it as a pure skill ranking confuses the player with the highest workload for the player with the finest touch.
The model still cannot see the pass
Expected assists is only as good as the underlying chance model, and that model was trained to value shots, not passes. It does not know that a cutback was disguised, that a pass was threaded on the half-turn under a press, or that the passer had to bend the ball around a defender’s outstretched leg. Two assists to the same square of grass—one a routine lay-off, one a pass that beat three players—carry the same xA if they set up the same shot. The very artistry that separates elite creators from tidy ones lives in the pass, and the pass is the part xA does not evaluate. It is measuring the destination and ignoring the journey.
Why it still beats the raw assist
For all that, xA is a clear upgrade on the assist, and treating the two as equally flawed would miss the point. The assist confuses a creator’s output with his finishers’ form, swings violently on tiny samples, and can vanish for a brilliant provider stuck with wasteful teammates. xA removes that finishing noise and stabilizes far faster, which makes it much better at telling you who has been creating good chances regardless of whether they were buried. It correctly refuses to blame a playmaker for a striker’s missed sitter. The critique here is not that xA is bad. It is that xA answers “how good were the chances my passes created” and gets read as if it answered “how good is my passing,” which is a related but different question.
How to read a creator’s number
Use xA as a chance-creation stat, not a passing grade. Put it next to key passes and to xA per shot assisted, which strips out volume and reveals whether a player’s service is genuinely high-quality or merely high-frequency. Watch the passes themselves, because the difficulty xA cannot capture is exactly what your eyes are for. And separate the creators who live near goal from those who create from deep, since the metric flatters the former. StatLine’s Premier League player tables carry the assist, appearance, and attacking lines that let you set a headline xA figure against the volume and role that produced it, so the number informs the judgment instead of substituting for it.
The pass before the pass disappears
xA also stops at the final ball, which erases much of the buildup that actually created the chance. The disguised through pass that springs a teammate to square an easy tap-in earns nothing, while the teammate who rolled in the tap-in collects the xA—the same distortion the secondary assist was invented to address, reappearing in a modern metric. Football’s best chances are frequently two or three passes deep, and the decisive action is often the penultimate one that unlocked the defense rather than the tidy final touch that beat the last defender. Because xA credits only the pass directly before the shot, it systematically overlooks the deep-lying playmakers and line-breaking passers whose work sets the table without appearing on it. The creator who does the hard part one pass early is invisible to a metric that only looks at the last one.
The honest read
Expected assists is one of the best creativity metrics football has, and it fixes a real defect in the assist by refusing to punish a passer for his teammates’ finishing. But it works by grading the shot rather than the pass, which means it rewards proximity to goal and volume of service more than the passing skill its name evokes. Read it as a shot-quality-of-chances-created stat, strongest as a stable, finishing-proof measure of output and weakest as a verdict on artistry. The best pass of the weekend is often the one that set up a hard shot from a bad angle—and xA will barely mention it.