3D player pose rarely improves football action value predictions
Does 3D Pose Improve Action Valuation in Football? An Empirical Study of Pose Representations and Learning Objectives
Computer Vision and Pattern RecognitionMachine Learning
Summary
Football teams try to figure out which players help score or prevent goals. These researchers tested if adding 3D body posture information to prediction models helps estimate player value better. They found that using 3D pose often did not improve predictions, except in a few cases related to body orientation. However, including pose did help predict what action a player would take next. So, 3D pose helps understand immediate moves but doesn't often improve estimates of overall player impact on scoring.
What this means in practice
- •For sports analytics teams: Improve short-term action prediction in football by including 3D pose data of players to better forecast immediate moves.
- •For sports performance analysts: Assess the limited impact of 3D pose data on overall player value models, focusing instead on refining training objectives for better predictions.
Tested on one dataset.
Authors
Rikuhei Umemoto, Kenjiro Ide, Soham Majumder, Ulf Brefeld, Keisuke Fujii
Abstract
Football action valuation estimates players' contributions to future scoring. Existing multi-agent deep reinforcement learning (MADRL) models use 2D positions and velocities, omitting posture information relevant to imminent actions. We investigate whether adding 3D pose to a SARSA-based model improves observed-action values relative to attack returns: goals, conceded goals or final expected possession value. Using UEFA Champions League 2022/23 event, tracking and 3D pose data, we compare six representations (no pose, raw joints, torso and head orientation, and spatial-temporal graph convolutional network (ST-GCN) embeddings) across five learning settings and three seeds. The settings vary temporal-difference targets and action-prediction loss, with Monte Carlo return regression as a reference. Across 50 on-ball and off-ball comparisons, pose robustly improved return prediction in only three, all using torso and head orientation, while worsening it in twelve, mostly with ST-GCN. Training choices had greater effects: removing the action-prediction loss raised on-ball skill of position-only values from 0.003 to 0.040 and their correlation with expected goals from 0.23 to 0.60; the improvement in return prediction persisted at matched training length. Separately, pose improved prediction of the on-ball action 0.2 s ahead (macro AUC 0.914 to 0.934 with raw joints), while the passer's pose during execution improved pass-success prediction (AUC 0.844 to 0.856). Within this training protocol, pose informed imminent actions but rarely improved observed-action values.