In this letter, we investigate the pilot contamination problem in massive MIMO networks from a system level point of view. We propose two game-theoretic approaches that model the pilot assignment problem. We also prove that one of these games is a potential game, and show that the adaptation processes following best and better response dynamics will converge to a Nash equilibrium. We then model the problem as an optimization problem. Finally, we compare the game theoretic results with the optimal and random pilot assignments. Our simulation results show that the game solution significantly outperforms the random assignment and performs as well as the optimal pilot assignment.