Go wasn't "solved" with MCTS. They ignore that there is caching of good moves (a good move 5 turns from now is probably good now too). In StarCraft, the difference between moving 36 degrees to the left doesn't make a big difference, so there are many move combinations which will provide nearly identical output. Additionally, alpha go did use an evaluation metric to measure the board value. Evolution of strategies just sounds like generic algorithms all over again