Towards Automated Football Analysis: Algorithms and Data Structures

Analysing a football match is without doubt an important tas k for coaches, clubs and players; and with current technologies more and more match data is collected. For instance, many companies o ffer the ability to track the position of each player and the ball with high acc ura y and high resolution. Analysing this position data can be very useful. Nowadays, some companies o ff r products that include simple analyses, such as statistics and basic queries. It is, however, a non-t rivial task to perform a more advanced analysis. In our research, we assume that we are given only the position data of all players and the ball with high accuracy and high resolution. In this paper we present two to ols. Our first tool automatically extract (from the position data ) a list of certain events that happened during the football match. These events include kick-o ffs, corner kicks, passes etc. In experiments we could observe that our method is very fast and reaches a high level of correc tness. We also learned that errors in the event detection are hard to avoid completely, when looking at only the position data. Our second tool aims at analysing a single player’s trajecto ry (the sequence of all positions during a game). More precisely, we look for movements of a player that are rep ated often (so called subtrajectory clusters). For example a left wing attacker runs from the centre-line al ong the left side of the field towards the opponent’s goal. And this attacker might repeat this type of mo vement very often during a game (or perhaps multiple games). Our goal is to detect this kind of frequent m ovements automatically. Experiments showed that this method is computationally expensive. Neverthele ss, it reliably identifies subtrajectory clusters, which then could be used for further analysis.