This paper extends the forward Monte Carlo (FMC) method, which have been developed for the basic types of American options, to the valuation of two-asset American rainbow options. The main advantage of this method is that it does not use backward induction as required by other methods. Instead, the proposed approach relies on a wise determination about whether a pair of simulated stock prices has entered the exercise region. A series of numerical experiments are provided to compare the performance with the binomial tree model and least squares method and demonstrate the efficiency of the forward methods.