这里有bug，最好是做一次，N次都取[0]的加和不是1

[Green-19]

ChineseChess-AlphaZero/cchess_alphazero/agent/player.py

Line 304 in 7f45b0c

p_ = (1 - e) * p_ + e * np.random.dirichlet(dir_alpha * np.ones(move_counts))[0]