Author: Uri Blass
Date: 07:11:10 12/09/02
Go up one level in this thread
On December 09, 2002 at 09:55:50, Robert Hyatt wrote: >On December 09, 2002 at 06:41:25, Uri Blass wrote: > >>On December 08, 2002 at 21:18:47, Robert Hyatt wrote: >> >>>On December 08, 2002 at 13:40:55, Dieter Buerssner wrote: >>> >>>>On December 08, 2002 at 12:58:53, Uri Blass wrote: >>>> >>>>>I think that Crafty or yace are better than every engine without tranposition >>>>>table. >>>> >>>>Interesting comment. I wonder, how you came to this opinion. It may mean, that >>>>something in the transposition table implementation of my program goes really >>>>wrong. >>>> >>>>Regards, >>>>Dieter >>> >>>I think that in endings, a non-transposition-table program will get out-searched >>>so badly it will lose almost every one. Try fine-70 with and without to see >>>what >>>I mean, as a worst-case example... >> >>I believe that if you add programs the right knowledge they can solve fine-70 by >>search without transposition table. > >I'm not convinced. It takes 23 plies to win the pawn. That is tough with no >transposition table - try it... I think that even more than it and only in qsearch after 26 plies white wins a pawn. The right knowledge is needed for it and programs do not have it. I think that you first need to do table of distance to pawns in order to detect by static evaluation if one side can win a pawn. The idea is that if one king is closer to a pawn than the opponent then the side can win the pawn. after knowing this you may also decide about partial extensions for moves that get closer to the target pawn when the distance is the same. > > > >> >>If they try to search 26 plies brute force search with no pruning and no >>extensions then I expect them to have problems. > >If they _do_ prune the chances are high they will prune the wrong move. In a >position >like this one, what looks bad can actually be winning... I agree that some ideas of pruning may not work here but there are still ideas that may work. You can prune lines when white win material because if black does nothing and let white to win material the exact value is not very interesting. > > >> >>I do not know about a single program with the right knowledge to solve fine 70 >>without transposition table but it does not mean that there is no program. >> >>Uri > >Some have tried. Monty Newborn had a pawn-endgame-only program called >"peasant." It >couldn't do this one either... The hardware at that time was also slower than the hardware of today and I believe that a smart pawn endgame program should solve it(programmers simply do not spend a lot of time on pawn endgames because they do not happen very often in games. Uri
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.