Simulating human grandmasters: evolution and coevolution of evaluation functions