All this is very difficult to implement in chess, and in practice, the method has not been developed enough to create a machine that could play chess as a master. Checkers are easier task. Equivalence figures significantly reduces the number of combinations examined. Also, partly due to this homogeneity, the game much easier checkerboard divided into separate stages than chess. Even in a checkers endgames main task is no longer to take shape and to be in contact with the enemy, achieving a position to take shape. Evaluation of moves in the game of chess should be made independently at different stages. Not only the ending is different from the most important relationship in the middlegame, but also in the opening extension figures in the position providing freedom of movement for the attack and defense, is much more important than in the middlegame. Therefore, we can not even approximately satisfied with a uniform assessment of different weighting factors for the game as a whole, but must break the learning process into a number of separate stages. Only then can we hope to create a learning machine that will be able to play chess as a master.