What can be measured from the broadcast? Validity of single-camera video for football-kick kinematics and swing-leg joint loads

This method-comparison study examined the agreement of single-camera broadcast-like estimates of football-kick kinematics and swing-leg joint moments with optical motion capture.Reference data comprised seven ball kicks from two CMU participants and 83 MoVi air kicks.A virtual camera varied view, player size and frame rate, using detector residuals measured on 200 Premier League shots.Realvideo validation used 162 recordings from 82 MoVi participants; a digital twin tested the simulation.Six kinematic features and three model-derived joint moments were assessed using absolute error, agreement and exploratory operational criteria.Broadcast detector error was 7.6-12.9% of player height.In the typical simulated condition (80 px, 25 fps, known contact and replay views), no single-camera variable met the selected criteria; the highest ICC was 0.64.The digital twin predicted aggregate real-video errors (r = 0.84).Three simulated replays with laboratory bias correction met the ranking criterion for kicking-knee flexion (ICC 0.80) and the laboratory-comparison criterion for trunk leans.A load index correlated with laboratory ankle speed (Spearman rho = 0.79), but its broadcast agreement was poor (ICC 0.09).Adaptation improved unseen air kicks but worsened CMU ball kicks.Thirteen smartphone recordings provided an illustrative field example without a reference measurement.These findings support cautious benchmarking and further football-specific validation, rather than individual clinical, scouting or injury-risk decisions from current single-camera estimates.