Building Chess Endgame Databases for Positions with many Pieces using A-priori Information

This note describes a possible solution to some endgames in chess with many pieces, when a complete retrograde database construction will fail because of the size of the problem and the time needed to build it up. The note shows how human knowledge, position-specific limitations and existing databases can be combined to prove a position to be won, drawn or lost, and providing the game theoretical optimal moves. The approach presented is fast enough to give results within minutes to hours when a normal PC is used. The computations with our program “Endgame Freezer” on the sample position Shirov vs. Fishbein in Section 3 took about eight minutes to complete, for example.