Expected Assists

The same shot value, credited to the passer

xA and xG price the same thing: the probability that a shot becomes a goal. xG asks how likely the shooter was to score. Expected assists asks the same question, then hands the number to the pass that created the chance instead of the player who took it.

The inputs are the ones that already build xG on the shot itself:

• Distance and angle. A chance closer to goal, and more central to it, is worth more regardless of who supplied it.
• Body part. A pass that sets up a clean strike is worth more than one that forces a header under pressure.
• Assist type. A cut-back to the penalty spot beats a cross from deep, because it leaves the shooter a cleaner sight of goal.
• Defensive pressure. How exposed the defence was when the shot went in.

Run those inputs on the shot that follows, and the result is that pass's xA. A ball into a crowded six-yard box scores low even when it looks skilful. A simple pass that finds a striker unmarked twelve yards out scores high, because the shot it created was always going in more often than not.

One rule sits above all of it: the pass has to produce a shot. No shot, no xA, however good the ball looked. Expected threat is the metric that rewards progression without a shot attached. xA does not.

Three ways to read the number

A creator let down by finishers. A winger delivering good chances match after match but registering only a couple of assists all season is doing their job. The passes are finding dangerous positions. The finishing at the end of them is not landing.

xA settles faster than assists. An assist needs somebody else to score, so a short run of matches can flatter or bury a player depending on who else was on the pitch that day. Strip out the "did it go in" step and the number stabilises across far fewer matches than the raw assist column does.

Two wingers, same assists, different pictures. One delivers early crosses from deep that occasionally get finished. The other consistently finds cut-backs onto a teammate's strong foot. Same assist tally, very different chance quality supplied. xA is what tells them apart.

xA settles the argument assists cannot

Raw assists reward the passer for somebody else's finishing. Expected assists asks only whether the pass created a good chance, which is the part actually within the passer's control. A high xA total sitting well above the assist count is not an argument against the creator. It is an argument for better finishers around them.

Three misreadings, corrected

• "xA is the same as key passes." A key pass count treats every pass leading to a shot as equal. xA weighs the quality of that shot. A low ball squared into traffic counts for far less than a pass that sets up a tap-in, even though both register as one key pass.
• "A high xA guarantees more assists soon." Only if the finishing improves. xA measures the chance supplied, not what happens next. A team stuck with poor finishers can keep generating high xA without the assist column ever catching up.
• "xA measures the passer's skill alone." It measures the chance created, and whether that chance is converted still depends on the shooter. Two identical passes can produce very different outcomes for reasons that have nothing to do with the pass itself. That is exactly why comparing assists against xA is useful: the gap belongs to the finisher, not the passer.

Combine xA with goals scored via expected goal involvements for the fuller attacking picture in one number.