Author: Uri Blass
Date: 07:27:19 08/31/02
Go up one level in this thread
On August 31, 2002 at 09:56:52, Joachim Rang wrote: >On August 30, 2002 at 14:45:43, Arturo Ochoa wrote: > >>Hello: >> >>I post now the final results for the Match Yace + My Modified Book Vs. Yace + >>NoBoook. >> >>I haven't analyzed the games but the result is clear: 13.5 - 6.5 for Yace + My >>Modified Book. >> >nice test, but 20 games say nothing. Play at least 100 or 200 games. Then on can >estimates the influence of book. 1)The result is 12-7 if you adjudicate loss on time correctly. 2)It is more easy to suggest other to play 100 or 200 games. From looking in the games I do not consider book as very important. My opinion is that if book could beat no book only 12-7 then book is expected to get even worse result against a small book only to get the opponent out of book. I think that it is better for programmers to test without book because testing with book may hide weaknesses of the program and the target should be to improve the engine. From looking in games I can see that most of the games were not decided because of book. Example: Yace lost one of the games because of overevaluating passed pawns when most programs have no problem to avoid Rxf6. [D]6k1/5p2/ppr2Bpp/2n5/PQ3P2/4P2P/6PK/2r5 b - - 0 45 am Rxf6 I believe that even the games that were decided because of mistakes in the opening are important for improving the knowledge of the program. Example from game 19 [D]r1bq1rk1/pppp1p2/4pn1p/4Nn2/1bP2Qp1/2N3B1/PP2PPPP/R3KB1R b KQ - 0 11 I believe but not sure that Bd6 was a bad move and d6 could give black at least equality. I admit that it is better to look at database of games and analyze before deciding about it. I believe that better knowledge of the opening stage can help to find the right move here. Uri
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.