Chess is far too complex a game to be solved. There are literally so many possible combinations of moves that the number cannot be written down except in scientific notation, there are 64 possible squares and each square can have either a white or black piece and that piece can be either a pawn, a rook, a knight, a bishop, a king or a queen, the fact there can be only a limited number of peices on the board only adds to the processing time which is why you can setup a board however you like and older engines won't bat an eyelid as its simpler just to evaluate all possibilities, now you cross multiply all those factors and then multiply by the possible number of turns, including repetitive moves which end in a forced draw (otherwise there are infinite possible games). No computer existing today, even the most powerful super computer, is capable of solving chess, this is a fact - heck they only recently solved checkers which only has 32 squares and only two types of peice, one of which can only move to two squares (or one if they're forced to take). The best engines can do is evaluate something like 20 moves deep and even then you'll be tapping your foot waiting for the thing. They play using a book of predefined openings which they follow provided the opposition uses a conventional opening/ response. Once the game inevitably becomes unique it switches to a tree based solver that has to evaluate all possible moves, hence the wait and the very limited depth. I used to be able to see 5 or 6 moves ahead but this is not because humans evaluate all possible moves but because they evaluate the most likely moves only, and at the higher levels you cannot fool your opponent, you have to outwit them by creating a series of moves that ends in the unavoidable checkmate of their king, often due to a seemingly innocuous move in or soon after the opening. The best modern chess engines don't even try to do this anymore because not only is it impossible, it's also really inefficient. Heuristic based problem solving is far more effective for problem spaces with such huge fields, and even more so for continuous problem spaces where the tree is branches so quickly. Neural networks are the best heuristic problem solvers we know of, I mean the same principals are how all animals learn, including humans. These newer engines wipe the floor against older tree based solvers. I've played around with them in unity and I can tell you they're extremely easy to implement and they solve many problems you'd otherwise need to dedicate a lot of resources solving in a more conventional manner. This same technology is now beating professional dota 2 players.
