Author: Reinhard Scharnagl
Date: 03:30:02 01/28/04
Go up one level in this thread
On January 27, 2004 at 16:52:59, Mike S. wrote: >[...] >You always argue like engines would be complete unable to play openings >themselves. That is not at all true. I state that using huge libraries during the phase of opening is contraproductive to gain improvements in detail evaluating functions, which could then better supply engines in the beginnings of chess games. >[...] >So why expect from chess engines, >that they find all this correctly in 3 minutes?? :-)) You are describing an ultamitive goal. If you deny, that it could be reached completely, why also skipping any possible progress in programming? >In general, engines will be better than IMs and GMs anyway, when "normal" >(normal for computers means very deep) tactical things have to be calculated in >the opening. Im not talking about Kasparov, but "normal" GMs. Most engines know >the common opening principles quite well (different quality of engines >undisputed). May be computer can handle shuffle chess openings better then human GMs, but this does not mean, that computer would handle it optimal. >[...] >I'd like to hear a comment about that from you. It seems to contradict to what >you always tell about engine's weaknesses without opening book. >[...] I would like to know, how you can decide, whether a chess program uses or includes opening knowledge or not. In my "fair chess" proposal I am demanding a size limit for engines at 1/4 MB (compressed, see at my homesite). Most detail evaluating functions have been optimized in its parameters by analysing myriads of chess games. This finally simulates an understanding of openings, which may be only slightly existant. You will see this, when such programs try to handle FRC starting positions. Though such programs might have good success in normal chess games, that does not mean that they would be able to navigate similar successful through exotic positions. Optimizing detail evaluating functions that way therefore may give a respectable amount of Eló, but it is leading into a dead end, because it generates no real understanding of positional play. And because the existance of centuries old opening knowlegde has been evil to computer algorithm development by trying to simulate or implement huge opening libraries, the way out of this dilemma is to enable programs to handle still less explored fields like FRC or Capablanca Random Chess, and to LIMIT the SIZE of chess engines rsp. their persistant storage. Regards, Reinhard Scharnagl
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.