Author: Mike S.
Date: 23:21:32 09/14/01
Go up one level in this thread
On September 14, 2001 at 19:09:40, Roy Eassa wrote: >It's hard for me to understand calling a game a draw when in reality it is a >win. The tables do not take sides -- they reflect the actual truth of the >positions in question. A fully pre-announced forced checkmate in 51 (w/o pawn >moves or captures) is now to be a draw?? I think that the 50 moves rule reflects the tradition of chess being a game. While we are used to think that a (tablebase-)position is "simply" either won, drawn or lost, in practice there is a bandwidth ranging from more or less clear wins, like a successful mate attack, to i.e. difficult technical endgame wins like some KRB-KR etc. Very long wins, which may take up to several hundred moves, are that narrow to a draw (and very far away from a "normal" game practice), that they are declared a draw by the 50 moves rule. They are just not won "enough". It also has a practical sense. A player (who is no computer) may have a position won in 30, but may not be able to play perfectly. The game has to stop somewhere. The 50 moves rule is ok IMO. If programmers can improve tbs usage in such cases where the win needs too many moves, they will probably try to find kind of 2nd best moves, i.e. continuations where the defender will have to find "only" defense moves often, or be mated before the 50th move if he fails finding them. This could lead to interesting attempts to win drawn positions, even those which are a draw for other reasons. Of course this idea would only be effective against humans, or against programs which don't use tbs. Btw. I think, and expect, that usage of all available 5+ piece tbs will be reconsidered by program authors, as huge experiments have shown, that in practical positions there is no advantage visible. Sometimes, engines without the tbs were even more successful. Ups and downs (speed) have been discussed often I think. Late M-Chess Pro versions allowed to define tbs directories which were used differently: One for access within the tree search, and the other one only when a database position already was at the board (I think programmers call that "at the root"). Good idea IMO, but nobody copied it so far. M-Chess uses Edwards bases though, which are more memory consuming AFAIK. Maybe that was the reason, but I think it is also considerable with Nalimov tbs (while another method, the Nimzo compressed data - NCD - tablebase info seems to remain either unknown or not an attractive enough idea for most other programmers). Regards, M.Scheidl
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.