Reinforcement Learning on a Futures Market Simulator