Human interaction can be viewed as a co-ordinated and synchronized system of behavior. The present type of analysis corresponds to a view of behavior organization as the repetition of a particular type of intra- and inter-individual (hierarchical/syntactical) behavioral patterns. Each of these patterns (called T-patterns), is characterized by significantly similar time intervals between its components over pattern occurrences. These patterns have a hierarchical structure, as each pattern component can be a simpler pattern of the same kind. Therefore, the pattern definition and the corresponding pattern detection method focus on relationships between the occurring times series of various behavioral event types. Such series also constitute the data for the pattern detection software, called Theme, which has been developed especially for the detection of such patterns in behavior. Theme has mainly been used for the analysis of intra- and inter-individual real-time behavior records with varying number of actors. It is able to work scale-independently with time resolutions of seconds, days, month’s etc. and with any kind of behavior. Theme can be used alone but using an especially developed interface it can also read and analyze the relevant types of observational data files (.ODF) created with The Observer.

To demonstrate this approach two studies are presented here, a soccer game and a boxing game analysis. The soccer game was coded with The Observer using categories such as player and ball positions and key game events such as goals, shots on goals, corner kicks, completed passes, intercepted passes, ball possession, etc. The boxing game was coded with Theme using categories such as position of boxer and relevant movements, such as left hook, right hook, knock down, etc. Results, from both studies, are demonstrated, both graphically and statistically, and examples of detected patterns are displayed and related to the results of the soccer game and the boxing game. The new Theme -- The Observer interface will be demonstrated to Observer users.

The results already suggest the possibility of discovering new kinds of profiles for both individuals and teams using the detected behavioral patterns in combination with elementary statistics. Moreover, some answers are already suggested to questions like: "Are there certain patterns that are related to doing well?", "Are there certain patterns that could be labeled bad?" or "What responses seem to be evoked by certain actions or sequences of actions?". Coaches could possibly use this kind of structural information when selecting players or when searching for the opponent’s "weak spots".

For more details about this work, click here.