We apply the recurrent reinforcement learning method of Moody et al. (1998) in the context of the strategic asset allocation computed for sample data from the United States, the United Kingdom, and Germany. It is found that the optimal asset allocation deviates substantially from the fixed-mix...