Author: Bob Durrett
Date: 11:41:25 12/23/02
Go up one level in this thread
Dann's comments below about closed positions are very interesting. Perhaps it is possible to write a chess engine, or a subprogram within a chess engine, to excel at solving closed positions, even if the engine cannot "become aware of" the fact that the position being examined is "closed." On the other hand, if a specialized "closed positions" subprogram were to be called, then it would be necessary for the program to "know" when to call that subprogram. There seems to be a tradeoff between the size of a program [measured in lines of source code???] versus the number of specialized subprograms contained in the overall chess engine program. Hypothetically, there could be hundreds or millions of specialized subprograms, one for each different kind of chess position. But that would be impractical, of course. Crafty, at about 40,000 lines of source code, is surely about as big as anyone "in their right mind" would want their chess engine to be. [Maybe 40,000 is too much too!! : ) ] Bob D. On December 23, 2002 at 14:25:51, Dann Corbit wrote: >On December 23, 2002 at 14:14:13, Bob Durrett wrote: >[snip] >>So, if we can put these obvious observations aside, what else is there about the >>initial position that is "special"? Another question is: "Do chess engines do >>as good a job of analyzing the initial position as they do for other positions?" > >No. For the most part, they do fairly poorly, even though good chess engines >will cough up e4, d4, Nf3 or some other decent start for the most part. > >>One thing is that most of the pieces are poorly positioned. Mobility is bad. >>Tactics are missing. Generally it's not a good position. > >It's a very even position. If I am not mistaken, the only advantage for white >is tempo. > >>OK. Now, do you suppose other positions occur in chess that are as bad as this >>one? If so, how well do chess engines do with such positions? > >Chess engines do badly with closed positions. They do worse with closed and >locked positions. Hence, they try to avoid them. > >>Perhaps chess engines are not optimized for analysis of such positions because >>they occur so rarely. No need to worry about the initial position, since the >>book takes care of it. > >Closed positions are neither rare nor frequent. They are a choice. > >>A big question in my mind is whether or not chess engines do a reasonable job of >>analyzing the initial position. Given the absence of an opening book, is the >>initial position really all that difficult for an engine, when compared to other >>positions? > >The Evans Gambit is a nightmare for chess engines to analyze. Hence, most chess >engines will have a ?? at some branch point to avoid it. > >>Anyway, I could go on and on with such questions. >> >>What do you think? > >It is well known that chess engines have a spot of trouble with closed >positions. Even at that, they will still beat 99.9999% of the people who play >against them. > >I think a more interesting question than: >"What positions do chess engines have trouble with?" >is: >"How do we fix it?" > >There are people who have written recognizers for closed positions. But that is >not the same thing as writing a solver for them. Not even close.
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.