Author: Nouveau
Date: 06:33:36 12/23/98
Go up one level in this thread
On December 23, 1998 at 08:32:39, Harald Faber wrote: >On December 23, 1998 at 08:23:33, Nouveau wrote: > >>> >>>This reminds me on the Nunn test which indeed is a thematic test. 10 (equal) >>>opening positions are taken out of the opening theory. >>>Indeed some try to test strength with this test... >> >>I understand the Nunn test as a "engine without opening library" test : it's a >>try to compare relative strength of engines. I think it's a good point to >>evaluate "pure" strength. >> >>That's not the direction I thought. The idea is to look at the way programs deal >>with different openings : some are complex, others positional or strategic, and, >>for the games I looked at, different programs have completly different evals for >>these positions. > >The Nunn test has different openings. Could you specify your idea where the >difference is? First, the Nunn test is without opening books, second the Nunn test has "only" 10 positions. Let's take an example to clarify : Theme : King's gambit (I'm an afficionado ;o) Players : Hiarcs7, Fritz5.32, Rebel10, Junior5, MChessPro7, Nimzo99, CM6000, Genius6 They play a multiple round-robin (white and black for each starting position) Positions : - Fischer defense (1.e4 e5 2.f4 ef 3.Nf3 d6) - Kiezeretsky (1.e4 e5 2.f4 ef 3.Nf3 g5) - Modern defense (1.e4 e5 2.f4 ef 3.Nf3 d5) - Bishop defense (1.e4 e5 2.f4 Bc5) - Counter-gambit (1.e4 e5 2.f4 d5) And with time, why not sub variations ? For this example, we have, for each program 35 games with white and 35 with black that is a good set to discover strengths and weaknesses for this opening. Maybe (I insist : _maybe_) would we discover a true King's gambit killer or a King's gambit specialist !!! It would be interesting for human player (for trainig or studying). What do you think ? > >>Besides it would be interesting, for me (am I alone in the case ?) to know which >>programs play better (understand better ?) the King's gambit, the Schliemann >>gambit of the Ruy Lopez or the Fajarowicz gambit. Other may be interested in the >>Sicilian Dragon or the Giucco Piano. I'm sure we would find great difference >>between programs of a generally same strength. > >I am sure but that is what opening book responsibles for commercial chess progs >do. They try to find out which openings are bad and which are good for the >program. Does that mean that progs only play openings they "understand" ? They would never play the French or the King Indian then ;o)). Besides most of the lines end with += or = positions that are to be played...well played ! > >>I read once (was it you ?) that the French is a bad opening choice for computers >>maybe could we find one that has better results than others with this defense >>? > >Such a difference will always be found in each opening. But I think French and >Kings Indian are 2 openings that won't be played well at least for the next 5 >years. So we do agree ! It would be interesting to investigate for important openings, for fun or study or trainig... Jeff
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.