Abstract
Hash signature collisions have been feared in computer-chess programs, dating back to the early Greenblatt program, and continuing until today. The fear has been that a signature collision, where the hashing function can take two different chess-board positions and produce the same hash signature, will produce errors in the tree search that will lead to gross playing errors and result in inferior play. As a result, chess programmers have spent significant amounts of time trying to reduce these signature collisions as far as possible (since a chess position cannot be uniquely stored in 64 or fewer bits, it is impossible to avoid all signature collisions). This research was done to answer the question: “Is it really worth all the effort to bring down the number of signature collisions to an absolute minimum?” The answer is surprising.