Okey is a four player social board game, similar to Rummikub, played by millions of people online or face to face. In order to play a session of the game, four people with similar expertise need to gather together at the same time. If a user leaves the game, users have to wait until a new player is found. Development of bots which can play similar to human players of different expertise levels is necessary for a better online user experience. In this paper, we formalize game playing bots which can mimic the behavior of a human Okey player. The bot needs to solve two main decision problems: which tile to include into the rack and which tile to exclude from the rack at each hand. We build a Design Matrix which quantifies the probabilities of tiles to form a group or run in the rack. We show that the two decision processes can be modeled using Singular Value Decomposition of the Design Matrix and hence we call the game bots that we develop the EigenBots. According to the experiments on over 5000 real games, our bots' behavior agree with the human behavior at least 80% of the time.