5

Are there any examples of positions where usage of the endgame tablebase gives an incorrect evaluation, due to the ignoring of the 50-move rule? At the moment I am not sure if endgame tablebases return the information specifically regarding the 50-move rule.

0
5

You are right that tablebases do not automatically contain information pertaining the fifty move rule.

To use tablebases accurately currently engines use the DTZ50 metric (distance to zeroing with fifty move rule) which tells you the number of winning moves until the 50-move count is reset. If this number is too big, the position is evaluated as a draw. Apparently this metric can give results that are one ply off.

To get perfectly accurate results these guys recompute tablebases with the fifty-move rule in mind. The resulting metric would be called DTM50 (distance to mate with fifty move rule). One of the differences between DTM50 and normal tablebases is that the winning line depends on the ply count and not only on the position on the board. Generally the higher the ply count, the longer the winning line, because a zeroing move has to be found within the next few moves, which restricts the possibilities. Here, you'll find an article about DTM50.

1

I've definitely heard of this, where an engine evaluated a position as winning right up until 49½ moves without a pawn move or capture, where the very next ply had to draw the game, and only then, evaluating the position as drawn. I don't remember the actual example that was given though.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.