Author: Mike Hood
Date: 11:18:30 04/07/02
Go up one level in this thread
On April 07, 2002 at 07:27:48, Sergei Smith wrote: >Which 5-man tablebases are used most ? >Some tablebases hardly ever get used and just take vauable disk space. It's not really a question of which tablebases are "used" most, it's a question of which are "needed" most. ie, If an endgame is difficult for a computer to win with its computational algorithms, there should be a tablebase to help it. For instance, KBN-KN will very rarely come about in real games, whether they are human-human, human-computer or computer-computer. However, if such a game does crop up a program that doesn't use tablebases will screw up with 99% certainty. On the other hand, the endgame KR-K (or KRP-K) is a common endgame, but a computer program will have no trouble finding mate without using tablebases. Another factor is that tablebases aren't just used when the endgame position is on the board. The 5-piece table bases are often used when there are 10 or more pieces on the board, if the program's search tree sees a path that exchanges pieces. Now let me attempt a very general answer: The tablebases Kxxx-K are less useful than the tablebases Kxx-Kx. There are only a few extreme positions where Kxxx-K isn't a win, and chess programs should be able to handle these endgames. Everything with pawns is very useful (KxP-Kx and Kxx-KP) The rest is more or less common sense. If it's an endgame you think is easy, the chances are that computer programs will share your opinion. For instance, the tablebases KQQ-Kx (even where x=Q) are less useful. The same is true for KQx-KN and KQx-KB. But having said this... if in doubt, keep all the tablebases. The complete collection of 3/4/5-piece tablebases is 7 GB, which isn't much considering the size of today's hard drives.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.