On the Numerical Approximation of Minimax Regret Rules via Fictitious Play