In this paper, we suggest an approximation procedure for the solution of two-player zero-sum stochastic games with continuous state and action spaces similar to finite elements and modified policy iteration approaches used for the solution of Markov Decision Problems.