Recursive Reasoning in Minimax Games: A Level $k$ Gradient Play Method