See Basketball Teams as Strategic Networks. The paper uses the first round of the 2010 playoffs to build network models of team dynamics. The graph nodes model players and edges represent ball movement between players. The paper then analyzes the properties of the resulting graph by computing degree centrality, clustering coefficients, entropy, and flow centrality. These properties allow the authors to characterize the games analyzed in terms of the graph structure. The authors conclude that the values of clustering (connectedness across players) and network entropy (unpredictability of ball movement) had the greatest correlation with team advancement. So it is indeed possible to erect a mathematical theory supporting the intuition often reported in the Blazer's Edge blog