Examining Online Poker as a Social Network

Data in the form of approximately 5 million No Limit Texas Hold’em hand histories collected from the online poker website PokerStars.com is used to model a game playing social network. Each player represents a node in the network, and weighted, directed edges represent the flow of money from the source to the target node, where the weight is equal to the amount of money transferred, normalized by stakes. Properties of the network structure such as degree distribution and weighted clustering coefficient are examined across different levels of stakes, ranging from $100 buy-in games (blinds of $.50/$1) to $1,000 buy-in games (blinds of $5/$10). We also explore the evolution of the network over time, creating probabilistic models for node arrivals and edge initiations. These models are governed in part by the unique set of incentives for edge creation, where each node seeks inward edges, but tries to avoid outward edges. We noticed several interesting patterns common to all networks across stakes, and attempt to utilize that information in our probabilistic model of edge generation. Despite our best efforts, the standard model of preferential attachment observed in class still performs best.