12,99 €
Problem: I have often observed an irritating behavior in variant analysis. It seems to be almost irrelevant which engine I use (e.g. Komodo,Stockfish, Shredder or others). I have observed the problem with Syzygy tablebases but also with other endgame databases.
The most striking feature is that in endgame positions (often near the "tablebase range") the engines give good results, but if later in the variation you come to a position from which e.g. with Kd5 the same position can be reached again, all engines show only this one move in the Kd5-variation, the variation stops here. This would not be so bad, because if I execute this move, the calculation continues properly. Much worse is that two or three moves before the possible recovery of the original position, it is apparently wrongly evaluated or probably simply falls under the table. I had some examples where definite draw (+/-0.00) is given, although the original position can be forced, previously judged with clear win. Since it seems to be engine-independent, it can really only be due to Chessbase or general engine weaknesses (perhaps related to tablebases) or my computer. I would be glad if you have an idea, maybe the problem is known.
Explanation: Many engines evaluate even two position repetitions as a draw and then stop the calculation process. Anything else would usually just be a waste of time, because computers "think" differently than humans. We are very aware of this and try to avoid the resulting breaks in analyses. Engine authors also try to avoid the grossest errors, but this is not always successful. The advantages of this implementation are much greater than these obvious disadvantages.
If you analyze with computers yourself, you should always consider this aspect and have it in the back of your mind. Tablebases are practically useless in many positions because there is too much material on the board.
Back to List
to send a comment please fill the form:
Name *
Email Address *
Subject
Feedback *
