In this paper, we introduce a model of a financial market as a multiagent repeated game where the players are market makers. We formalize the concept of market making and the parameters of the game. Our main contribution is a framework that combines game theory and machine learning methods. This approach allows us to consider markets on both a macro level, through game outcomes, and on a micro level, through the optimization efforts of players. Using simple equilibrium analysis, we show that our model explains situations where market outcomes are inefficient or unsustainable. We further apply our model to simulate market makers in the SP500 E-mini futures market and show that players learn to adapt their quotes to different market conditions.