In the present universe of current gaming condition bots are the intelligent agent that assumes a prominent job in the popularity of a game in the market. As these bots have gotten very unsurprising to the games. So here we are proposed an AI model for playing games with high level inputs using reinforcement learning. Algorithm works in the Atari Environment i.e. we are using 2D game. This model consists of the CNN (convolution neural network) for the inputs which is fully connected layers and find out the actions according to the inputs. In this learning -based approach, bots learned how to attack and ignore opponents so that bot can get maximum score. In this learning -based approach, bots learned how to attack and ignore opponents so that bot can get maximum score Then we tried the combine the input method which results maximum score of the bot in the environment for the better performance.