Author: Robert Hyatt
Date: 07:24:46 11/20/99
Go up one level in this thread
On November 19, 1999 at 18:13:31, Fernando Villegas wrote: >On November 19, 1999 at 13:23:46, Bruce Moreland wrote: > >>On November 19, 1999 at 03:08:35, Bernhard Bauer wrote: >> >>> >>>>>>> >>>>>>>Do I have a defective disc or is Nimzo 7.32 a defective program in this >>>>>>>area? Or have I missed something? >>>>>>> >>> >>>> >>>>There is no contradiction. >>>>you do not lose more than few elo ratings by not knowing KBN vs K and KBB vs K >>>> >>>>Uri >>> >>>This is the typical bean counters argument: Rating points. >>>You do not lose more than a few rating points by not knowing this and that, >>>so what you get is kind of a cripple chess with some nice features like the >>>insufficient material feature. >>>I want a chess program to be able to play chess in *any* possible position. >>>Every thing else is poor. I don't mind about rating points in computer chess. >>>Kind regards >>>Bernhard >> >>The programs are generalists, and that is the problem. There are some specific >>cases they do not handle well. KBN vs K in particular requires an extra >>heuristic in order to perform well. In a normal endgame position you will not >>try to force the opposing king to the corner that is the same color as the >>bishop, but this bit of information is extremely important in KBN vs K. >> >>So unless you program some extra help in, a program will play that ending >>sub-optimally. >> >>There are always going to be positions that the program doesn't understand as >>well as a person, since it is different from a person. >> >>You are right to demand that specific rare yet interesting cases be covered, >>though. >> >>This probably applies just as well in the middlegame. Computers play their type >>of game very well, and this is sufficient to do well in most middlegames, but >>there is obviously a lot of room for improvement. >> >>bruce > >I think I have said this time and again, but here I go again: why not to develop >more "fragmented" programs, that is to say, not a full program supposed to do >well in any aspect of the game, but one with a specific module to identify what >is going on and then select another specific module to treat it? I presume >something of he sort is currently done, but I suspect that a lot more can be >done. In fact, is what human beings does: we does not play chess with the same >package of ideas in the middle of ending, with this or that set of remaining >pieces. In our case is unconscious, we just change our parameters and to begin >with, pawns becomes a lot more important, the kind becomes ative, etc, but a >program could do so provided the super-module had a complete -or almost- table >of pattern positions to identify what is needed. >Cheers >fernando This can cause serious problems. There is a well-documented search/evaluation problem called "the discontinuity problem". If you suddenly switch from one engine to another, then at positions around this 'switch' you can make gross errors. Or you do like Fritz and have an eval of +1 before trading queens, and 0.00 after trading. The evaluation/search has to perform in a continuous way, slowly transitioning from opening to middlegame, and from middlegame to endgame. I used to have three distinct 'phases'. I am slowly eliminating those and making the program perform better. I don't believe it will be possible to beat GM players consistently with big discontinuities in the evaluation, or in the search space. They will notice the 'transition points' and take advantage of this. In ways detrimental to the program. It has happened to me many times. An example is king safety. You want to slowly phase it out as you approach the endgame, not just <wham> it is off. Because you might reach a position where your king shelter is gone, and you give up a pawn to trade away the piece that drops the king safety term to zero. And convert a likely win into a sure draw/loss. IE "think smooth" not "jerky"...
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.