Statistics for Towards Global Reinforcement Learning