ExtremeHunter vs RobustUCB numerical experiments from the paper Max K-armed bandit: On the ExtremeHunter algorithm and beyond: https://arxiv.org/abs/1707.08820