Evaluating the indoor football tracking accuracy of a radio-based real-time locating system

Nowadays, many tracking systems in football provide positional data of players but only a few systems provide reliable data of the ball. The tracking quality of many available systems suffers from high ball velocities up to 120 km/h and from the occlusion of both the players and the ball. Radio-based local positioning systems use sensors integrated in the ball and located on the players' back or near the shoes to avoid such issues. However, a qualitative evaluation of the tracking precision of radio-based systems is often not available and to the best of our knowledge there are actually no studies that deal with the positional accuracy of ball tracking. In this paper we close this gap and use the RedFIR radio-based locating system together with a ball shooting machine to repeatedly simulate realistic situations with different velocities in an indoor environment. We compare the derived positions from high speed camera footage to the positions provided by the RedFIR system by means of root mean square error (RMSE) and Bland-Altman analysis. We found an overall positional RMSE of 12.5 cm for different ball velocities ranging from 45 to 61 km/h. There was a systematic bias of 11.5 cm between positions obtained by RedFIR and positions obtained by the high speed camera. Bland-Altman analysis showed 95 % limits of agreement of [21.1 cm, 1.9 cm]. Taking the ball diameter of 22 cm into account these results indicate that RedFIR is a valid tool for kinematic, tactical and time-motion analysis of ball movements in football.