Author: Bob Durrett
Date: 06:29:47 10/23/02
Go up one level in this thread
On October 23, 2002 at 02:33:33, George Sobala wrote: >On October 22, 2002 at 16:37:34, Bob Durrett wrote: > >>On October 22, 2002 at 16:04:12, George Sobala wrote: >> >>>On October 22, 2002 at 15:06:40, Bob Durrett wrote: >>> >>>>On October 22, 2002 at 13:38:44, George Sobala wrote: >>>> >>>>>On October 22, 2002 at 09:41:18, James T. Walker wrote: >>>>> >>>>>>Computers will never "understand" speculative sacrifices. Just because Fritz >>>>>>thinks the move is bad is not proof. This sac might have worked against many >>>>>>humans who could not find the proper response. That's the nature of speculative >>>>>>moves in chess. The computer didn't get rattled where humans might have. If >>>>>>Kramnik could have computed the line to a forced win then it's not speculative >>>>>>and maybe not even a sac. >>>>> >>>>>I agree. Although computers are often touted as "tactical monsters" they have >>>>>some quite marked weaknesses in certain types of tactical positions and cannot >>>>>be relied on to prove anything! >>>> >>>>Not clear to me. Of course, I am not a professional chess programmer, so maybe >>>>I wouldn't know. >>>> >>>>But . . . >>>> >>>>What is there to keep a chess programmer from writing code that forces the chess >>>>engine to examine all such sacrifices? Yes, I know there would be a few >>>>technical difficulties. But, what about the essence of this idea? Is it so far >>>>out from what is do-able? >>>> >>>>Bob D. >>> >>>Brute force searching of all possible moves will always get the answer ..... >>>eventually. But the size of the potential search tree is what prevents computers >>>searching ALL possible moves. They take shortcuts. Now of course for the current >>>position they will look at all the moves, but if the success of the first sac >>>depends on an obscure move or unusual second sac a few moves later - they don't >>>see it as winning, or at least not for a LONG time. >> >>I don't see that it would be necessary to "search all possible moves." In the >>first place, only sacrifices would be considered by the subprogram [or it's >>equivalent]. Secondly, the number of moves could be reduced further by adding a >>few more appropriate criteria. >> >>Still looking for an answer since I don't see that the idea mandates "searching >>all possible moves." >> >>Bob D. > >Well if you don't search all possible moves there will always be some moves in a >position that you don't search. You must have used some criteria to reject these >moves. In certain positions, those criteria WILL be wrong, and you will have >rejected the winning move. I hope you know enough about chess to realise that in >complex tactical positions it isn't necessarily the sacrifices that are the >profound moves! This topic is interesting, so I would like to pursue it a little more. Disclaimer: My understanding of chess is at the USCF Class A level. Also, I am not a professional chess programmer. I don't play competitive chess anymore, having found more rewarding chess pursuits. But when I did play chess, I habitually always looked first at the sacrifices [for both me and my opponent], and tried to determine whether or not they could possibly make any sense. This helped to avoid surprises and minimized the chance that I would miss any pretty sacs. It also helped me to find combinations. When doing the "examination of potential sacs" activity, one's "mind set" is different. It's a different way of thinking. A chess engine's "mind set" is something that can be set by a programmer. It need not be the same at all times. It could change often during the game. Just like my wife, who claims she has the right to change her mind whenever she wants to. I don't see why a chess programmer could not write a program, or subprogram, to do the same sort of thing. It is clear, however, that the modern chess programmers are severely handicapped by "the force of habit." In other words, they suffer from the compulsion to rely on search algorithms. : ) But, maybe they could accomplish what I'm talking about without too severly offending the "God of Search Algorithms." Maybe they could build different "mind sets" into the search algorithms. Somehow. Bob D.
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.