Examples of complex systems, with time evolving connections between its elements, abound in the social, biological and physical domains. In many, if not most of these systems, elements are clustered in groups that also undergo changes with time. A temporal, clustered network can be an appropriate representation of such a system. We use the ”Shared Information Distance” [1], an entropic measure, to evaluate the change the network experiences over time. This measure can be split into two components: 1) the amount of change resulting from the evolving network cluster size sequences and 2) the connection activity between nodes. We study the usefulness of such a split analysis, guided by the intuition that different drivers may contribute separately to each component. While calculating the total shared information distance is computationally trivial if the network partition into clusters is known, finding the split is not routinely tractable, and we developed a heuristic to approximate it efficiently[2]. We apply our approach to the analysis of soccer games from the 2011 season of the English Premier League. Each game is modeled as a high-resolution (10Hz) temporal hypernetwork with simplices as clusters[3, 4, 5], parsed by player proximity. We find that the evolving simplices sizes are the major contributor to the overall information distance as the time progresses. We find also that the shared information distance decreases as the games develop, which suggests this measure may be a proxy for player physical exhaustion. Using our approach we also encounter correlations between spikes in information distance and significant events from match commentary at different time scales

--

Complex Systems,Networks,Association Football