Author: Robert Hyatt
Date: 07:18:47 09/12/04
Go up one level in this thread
On September 12, 2004 at 03:52:56, Sune Fischer wrote: >On September 12, 2004 at 00:42:50, Robert Hyatt wrote: > >>I do some of that. But the pieces can end up on squares where I normally don't >>want to push pawns... This is probably fixable, but not trivially. Other >>issues (IE g3 after O-O if white has a light-squared bishop to fill g2, but in >>FR that bishop might be elsewhere and not able to get to g2 easily. I saw lots >>of such stuff in testing so I just put the stuff aside... > >All of these terms can be generalized for use in FRC as well, I think the idea >is simply to check if there is a defending bishop plugging the hole in some way. Yes, but then I have special code for opening development. This code understands some of the normal pawn push ideas. IE in d4 openings don't block the c2/c7 pawn as it usually advances quickly. You need the same sort of stuff for FR positions as well, but the "pawns in question" can change. As I said, it isn't impossible, but the "stock evaluation" I have doesn't work very well with many FR positions. > >>>> >>>>How does your program do in 1. g4 openings? :) >>> >>>Relatively speaking no worse or better than other engines I hope. >> >>That's not what I asked. How does it do in g4 openings as opposed to normal >>openings? _that_ is the question. Why does it do badly? Would some eval >>tuning help? Probably. > >Well, playing matches can be seen as a means to and end to improve the engine. >If it does relative worse on g4 openings than other engines then that might >indicate some sort of weakness so I want to test that and find out about it. > >>>If everybody had a copy of those GM's at home I'm sure these matches would be >>>played :) >> >>Sure, forcing Kasparov to play English openings, forcing Karpov to play the >>Latvian, etc. Wouldn't be very revealing however... > >Indeed it would, every player should know his strong and weak points, perhaps he >plays the english better than he is aware of :) I'd bet the GMs _know_ their weak points. That's why they don't play certain openings. > >>> >>>The question of which engine to use for analysis is interesting in itself, >>>possibly it is even the primary question for most users, I know it is for me. >>>Books serve only to skew the picture here, IMO. >> >>I suspect there is _NO_ "best engine for analysis" any more than there is a >>"best human for analysis". I'd bet that each engine excels in certain parts of >>the game, and even worse, some excel against computers while others are more >>adept at playing humans. > >The problem is nobody can keep track of which engine is good for attacking, >which engine is good for defendig, which engine is good for early middle games >and so on. I'm not even sure how you would measure this, the only thing that >counts in the end is the average strength. Then someone is making decisions based on flawed information, which is not very good. All programs do well at tactics, and pointing out tactical mistakes. When it comes to positional concepts, it gets very fuzzy. > >>These "matches" don't show _nearly_ as much as many believe... > >They show me what I want to know, ie. how good is Fritz _without_ the killer >book from chessbase? Why does it matter? Why does it matter how Fritz does with a bad book? No endgame tables? Impossibly short time controls? No pondering? No learning? Why not test with "no code" as well??? >Suppose the book is worth 100 Elo and Fritz is the only one who is allowed to >use that book, now obviously Fritz will look 100 Elo stronger in all matches >than it really is, and obviously these 100 Elo are worth nothing to a >correspondence player who only needs the engine for analysis. Au Contrare, Fritz will be giving _good_ opening advice, for one thing... And if you expect _any_ program to give good advice on oddball openings, good luck... > >-S.
This page took 0.01 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.