
Pretty sure that this effectively does nothing as the point of using temperature is to skew the probability distribution to highest number, and to choose the move probabilistically, not choose the highest one. The point being that if the evaluations are very similar, we tend to explore rather than picking the highest.