Social network of the competing crowd

In this paper, we analyze the social network among competition participants at Kaggle.com. In particular, individuals, called members, are allowed to participate in various competition in teams. Each team may have one or more members. As a result, the relationship between teams and members may be represented as a bipartite graph. First, we project the bipartite graph into networks of teams, and members, respectively, and study their properties as one-mode social networks. We find that these social networks follow many of the typical characteristics of large-scale social networks. However, different from traditional social networks, given the competitive nature of the teams, we find a large number of connected components in these networks, which shows that players at this marketplace are densely but not widely connected. Then, we find meaningful statistics extended for bipartite networks, and compare them with such networks from other application domains. Similar to the literature, we find the need of investigating this raw graph in order to understand the relationship among players at the competing crowd better.