Possession value model

A possession value model extends shot-based analytics beyond the shot itself, assigning a value to every touch of the ball in a possession, passes, dribbles, tackles, not just the final shot, based on how much each action changes the team's probability of scoring before the possession ends.

These models, sometimes known by specific implementations such as VAEP or expected threat frameworks, divide the pitch into zones and estimate the scoring probability associated with having the ball in each zone and state; an action's value is the change in that probability from before to after it occurs. This differs from box-score counting stats and even from expected goals and expected assists, which value only the shot and the final pass, by crediting buildup actions like a defensive interception that starts a promising attack or a dribble that breaks defensive lines.

Analysts use possession value models to evaluate players whose contribution shows up in buildup play, defense and progression rather than goals or assists, complementing measures like progressive passes. The main pitfall is that these models are comparatively complex and less transparent than a simple shot count, are sensitive to the pitch zones and features chosen when building them, and different implementations can rank the same players quite differently, so results should be read as one model's estimate rather than a settled figure.