Author: jefkaan
Date: 08:02:45 12/04/05
Go up one level in this thread
On December 03, 2005 at 08:13:25, William Penn wrote: >The only purpose of these opening books, as far as I'm concerned, is to derive >statistical analyses for the different candidate moves. maybe in the beginning of the game you get an idea from the statistics, but later on statistics are useless, especially if they are based on historic human games with lots of mistakes. >Perhaps the statistics indicate which is the best move, well at least in the beginning, eg. after e4 e5 Nf3!! if use statistics to decide on sharp gambit lines, eg. in the KGA, after move 5 or so, you get insane results. Analyzing such a variant yourself, instead of statistics would get better results, but if you analyze a complicated variant, and try to find improvements with computer analysis, at least to avoid tactical errors, you need to either have the engine use its hashtable to 'remember' its analysis (like eg. Shredder 9 does). But also then, when there are lots of sidelines, and complicated endgame strategies, such an approach is far from perfect. So a 'mixed' approach of statistics and minimax, like the opening books of Arena and CA.8 can do, is better than statistics alone (like the Chessbase programs seem to do). or minimax alone (Bookup). But even then a problem with minimax (backsolving) is how to evaluate the end positions. GM evaluations, either from opening books like NCO, or specialized opening text books tend to give better evals than a simple engine. So in that respect CA.8 is likely to give the best results, as you also can use GM evals with their 'elite' opening lines. >It is not so easy. well i can agree with that.. :) but in any case with booklearning, a good engine, and playing lots of games, you automatically will get better opening books ! Jef http://superchess.blogspot.com
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.