Massive Multiple-Input-Multiple-Output (MIMO) is a promising technology to meet the demand for higher data capacity for mobile network in next generation communication system. However, due to the massive connectivity of mobile devices, the pilot contamination problem will severely degrade the communication quality and spectrum efficiency of massive MIMO system. As the blooming of artificial intelligence (AI), it provides a new vision to solve the traditional communication signal processing. In this work, we employ the core algorithm of AlphaGo, Monte Carlo Tree Search (MCTS) method to play the pilot allocation game. By converting to a Markov Decision Process (MDP) based game model, we employ MCTS as a reinforcement learner to solve the joint pilot sequence and power allocation problem. We map pilot-power allocation problem to a game based process by defined action, state and reward. MCTS will automatically find the optimal pilot allocation action with maximal reward value. Numerical results show that our proposed MPAS achieves a better CDF of ergodic spectral efficiency compared to the previous suboptimal algorithm.