Expected Assists and Goal Involvement: the numbers

Expected assists credits the pass that created a chance with the value of the chance itself. Combined with expected goals it produces goal involvement, a single figure that describes a player's entire contribution to chance creation and finishing.

How the credit is assigned

When a shot is taken, the model has already produced a value for it. Expected assists takes that same value and attaches it to the pass that immediately preceded the shot, provided the pass satisfied the provider's definition of a creating action. A player who plays the final ball into a chance worth a third of a goal collects a third of an expected assist regardless of whether the shot was converted.

The design intent is to remove the accident of conversion from the assessment of creation. An assist is only recorded when the shot goes in, which means the assist column rewards the passer for the finisher's accuracy. Expected assists rewards the pass for the chance it produced, which is the part the passer controlled.

Goal involvement: the natural sum

Adding expected goals and expected assists gives expected goal involvement, the combined value of the chances a player finished and the chances he created for others. For a forward who both shoots and supplies, this single number is a fairer description of his attacking contribution than either component alone.

The sum has an important limitation that is easy to forget: a player cannot assist his own shot, so the two components are mutually exclusive by construction. That makes the total clean, but it also means the metric cannot capture value delivered in the second or third pass before a chance, which is the territory of the more advanced contribution models.

Where expected assists misleads

The metric inherits every flaw of the underlying shot model and adds its own. It credits only the final pass, so a defender who splits a block with a line-breaking ball that a teammate then squares for a tap-in receives nothing, while the squarer receives most of the value. That asymmetry systematically rewards the last pass over the important one.

It also rewards volume in a way that flatters certain roles. A winger who swings in twenty crosses a match will accumulate expected assists from the occasional connection even if most deliveries are cleared, while a progressive passer who creates three high-value openings may record less. Reading the total without the count of creating actions repeats the same category error as reading goals without shot volume.

Set pieces and the inflated case

Set-piece delivery is where expected assists becomes most inflated. A corner taker who delivers fifteen corners in a match draws a small value from each contested header, and the accumulation can outstrip a creative midfielder's open-play figure without describing anything comparable in difficulty.

The disciplined treatment is to separate the phases: open-play expected assists, set-piece expected assists and the count of deliveries for each. Aggregated without that split, a dead-ball specialist and a creative playmaker can appear to be doing the same job when their numbers come from entirely different events.

Pairing it with the shot model

The natural companion to expected assists is the shot model itself, because the two share a unit. A player's expected goal involvement can be cross-checked against his minutes and his role: a central midfielder with a very low involvement figure is probably playing deep, and one with an unusually high figure is probably operating in the final third or taking set pieces.

The pairing also exposes the difference between creation and finishing. Two players with identical involvement can be entirely different: one takes most of his value from shot volume and the other from creating for others. Splitting the two components is what turns a single figure back into a description of a role.

Reading the columns together

The practical table is four columns: expected goals, expected assists, the sum, and the count of creating actions. Read across a row it describes a player's role; read down a column it compares like with like. Any comparison that uses only the sum will occasionally rank a specialist set-piece taker above an open-play creator, and the columns beside the sum are what reveal the difference.

For defenders and deep midfielders, involvement figures are naturally low and are not a criticism. The progression and build-up models cover that part of the game, and quoting a low involvement figure for a holding midfielder as evidence of limitation is a category error about what the metric is for.

Key reference points

• Expected assists transfers the shot's value to the pass that created it.
• An assist requires a goal; expected assists does not, which is the point of the metric.
• Goal involvement is the clean sum of expected goals and expected assists.
• Only the final pass is credited, which undervalues the pass before it.
• Set-piece delivery inflates the figure and must be reported separately.
• Read the components separately to recover the player's role.

Expected assists measures creation rather than reward, which is why it disagrees with the assist column and why that disagreement is informative rather than embarrassing.