Author: Sandro Necchi
Date: 10:34:18 09/04/05
Go up one level in this thread
On September 04, 2005 at 10:49:56, Jorge Pichard wrote: >On September 04, 2005 at 06:21:50, Sandro Necchi wrote: > >>On September 04, 2005 at 05:41:43, Jorge Pichard wrote: >> >>>On September 04, 2005 at 05:37:48, Jorge Pichard wrote: >>> >>>>Too many people like to use long GM's Openings for computer programs, but there >>>>is a drawback to this approach. Todays computer programs has reached a level of >>>>play that is higher than some of the best GM players. In order to avoid this >>>>from happening, most Opening Expert like Necchi and Harry Schapp to name a few >>>>will have to consider checking each opening where the two GM left as an even >>>>game with the top programs to analyse up to at least 6 more moves and only if >>>>after those 6 moves if the opening is still consider even, then the opening >>>>could be included with the rest of its opening library. We should be very >>>>carefull not to feed long Opening line above 12 moves from GMs. Please take a >>>>look a this game, and how many other bad human openings are still there in some >>>>of the best programs that we constanstly pit or match, favouring one side or the >>>>other. Here is a famous game where two of our best computer monsters were using >>>>the exact bad opening fed directly from a game of two GMs: >>> >>>PS: That is the reason why I limit Opening lines up to 12 half moves, specially >>>if those openings were played by human GM. >> >>Well, this is the simple way to avoid those problems, but not the best way to >>improve the book. >> >>Of course to work on the book extending and checking the lines takes a lot of >>time unless one can use the computer support...I have new ideas on this matter, >>so with the help of Stefan we should be able to see clear advances on this field >>too. >> >>Sandro >> > >Yes, it does take time to check the lines at the end of every openings with a >computer at least 6 moves deeper, but it is a safe way to avoid disaster and >every opening book which derived from human GM should be checked. > >Jorge I am looking for the best way to make the book better, not for the safer way. Also there are mistakes early too as some variations are not as good as one would expect from human games/scores. Sandro >>> >>>http://wired-vig.wired.com/news/print/0,1294,68243,00.html
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.