Expected Pass Value (xPV): A holistic framework for evaluating passing situations in soccer

Using pitch control and pitch value models to evaluate every aspect of passing situations.

Table 1

Resource	Date	Link
Tobias’ talk presented at the 2025 New England Symposium on Statistics in Sports (NESSIS)	2025-09-27	Talk Tobias

Overview

Passes are the most frequent events in soccer. Yet, to this day, analyzing passes statistically poses a substantial challenge. Existing metrics often lack context (e.g., successful passes) or are heavily outcome-dependent beyond the passer’s influence (e.g., assists, expected assists). In this work, we leverage spatio-temporal tracking data provided by PFF to gain deeper insights into passing. In particular, we develop a framework for valuing any real or hypothetical pass on the pitch based on two well-established components: (1) a pitch control model, and (2) a pitch value model. For (1), we derive a computationally efficient model to determine the likelihood of controlling a location on the pitch based on players’ velocity, movement direction, and angular mismatch to the location of interest. To estimate (2), we employ an XGBoost model based on a rich set of features that are derived from tracking data and capture situational context. Obtaining a value for every possible pass enables a holistic analysis of passing situations. More precisely, we can quantify the passing ability and decision-making of offensive players, and simultaneously the positional play of defenders. We apply our framework to obtain an expected pass value (xPV) for all passes from the World Cup 2022. Results show that xPV outperforms existing metrics, as demonstrated by higher correlations with players’ future performance indicators and market value. We further analyze defenders by their ability to prevent xPV and quantify offensive decision-making by comparing the actual xPV to alternative passing options.