Analysing performance in football is an important task for coaches and by the substantial develop-ment of technologies which have become a con-sistent component in sports, a huge amount of data is collected every day. This data can poten-tially provide information of the tactical or tech-nical performance, for coaches to evaluate. How-ever automated methods need to be developed in order to extract this information.
The purpose of this thesis was to develop and evaluate an automated algorithm to detect and classify passes in a football match using spatio-temporal data. The proposed method consisted of multiple steps including the application of suitable filters, pass detecting procedures and classifying pass based on length and direction which were detectable by the algorithm. The data consisted of one half of a football match played in the Danish Superliga and was collected by an optical tracking system. The results showed a substantial accuracy of the detected passes by the algorithm compared to the ground truth created by human observers (F-score = 0.80). In conclusion the method showed to be a useable tool for the detection of passes, however certain limitations regarding constraints in the algorithm and the data needs to be considered in future research.