Hand tune the regularizer
It is obvious that the MCTS only explores ONE path and allocates just one simulation to the others. So, it's not a surprise that the moves are just random.
Hand-tune that parameter so that each have at least a few sims.
It is obvious that the MCTS only explores ONE path and allocates just one simulation to the others. So, it's not a surprise that the moves are just random.
Hand-tune that parameter so that each have at least a few sims.