Reinforcement learning is one of the sub of machine learning. A machine learning agent learns from the feedback of the try-and-error in order to predict their next step. Machine learning can be use on various field and one of them is games. The challenge to win a game is that the player needs to come up with a good strategy. In order to produce good strategy, player need to play the game multiple time which are time, energy and money consuming. The objective of this research is to introduce a reinforcement learning agent in game that run the simulation of the game and produce improved results after each iteration. Then human can imitate the agent performance in order to improve their chance of winning the game. Reinforcement learning can be implemented in various method. This paper will focus more on Q-learning and State-Action-Reward-State-Action (SARSA) method. Both methods are chosen as both are almost similar except Q-learning is off-policy algorithm and SARSA is on-policy algorithm. The results of this paper is a list of results from previous research related to Q-learning and SARSA on different test field or setting. The second results are the proposed reinforcement learning methodology what will cover on understanding data, categorizing problem, finding the available algorithm and implementing the algorithm.