Fictitious play and Q-Learning was examined for the computation of equiilibria in Zero-Sum games. We tested convergence to Nash-Equilibria in several two-player(two agents) games, which are matching pennis, rock-paper-scissors and selling damaged goods. Regarding the approach that was followed, we tested fictitious play agent vs fictitious play agent(FP vs FP), Q-Learning agent vs Q-Learning agent(Q-Learning vs Q-Learning) and Fictitious play agent vs Q-Learning agent(FP vs Q-Learning).
The report is included here.