The FFG (Austrian Research Promotion Agency) funded project WiMoiS („Wissensbasierte Modellierung der individuellen Spielleistung“ – „Knowledge-based Modeling of Tactical Abilities in Game Sports“), which was conducted in cooperation with the SCCH (Software Competence Center Hagenberg, www.scch.at/de/) and the company Inmotiotec (www.abatec-ag.com/inmotiotec-rtls/) ends at December 31, 2016.

Aim of the project was the development of methods for the automatic recognition and assessment of collective and individual performances in game sports based on position data (tracking data) and machine learning. All necessary data therefore was gathered at the youth academy of the Austria Vienna Football Club (http://www.fk-austria.at/de/teams/akademie/info/). The processing of this data is coupled with a qualitative research process – see the following figure: Expert interviews are used to gather information how to analyse and assess team and player behaviour by knowledge. This information is used thereafter to model features based on the tracking data, which are used to feed machine learning methods for pattern recognition.

Object of the investigations are small-sided games in soccer. Data basis and input for the assessment system are position data of all participating players and the ball. In the meantime automatic collective and individual player assessments are possible for low complex test forms (e.g. 1 versus 1) and middle complex games (e.g. 3 versus 2), and a couple of analysis tools have been developed for high complex games (5 versus 5): recognition of passing strategies; pass sender and receiver evaluation; defence positioning assessment. Till the end of the project the association of those tools are promoted in order to enable comprehensive analyses for the 5 versus 5 small-sided game.

Kontakt: Dr. Roland Leser, roland.leser@univie.ac.at