Author: Dan Honeycutt
Date: 11:25:26 02/18/05
Go up one level in this thread
On February 18, 2005 at 09:01:08, José Carlos wrote: [snip] > > Hmm, does this make sense? I mean: > - if I'm testing killer book vs no book, I will win 100% of the games, just by >picking 2 won games (with white and black) and put the moves in my book. > - if I'm testing killer book vs random book not known by the book cooker (big >enough to not repeat lines in a match), it's impossible to prepare killer lines >for every possible game. In a match of 100, maybe 5 can be killer lines. That >can't make for 700 elo points. > - if the book cooker knows the random book in advance, we're near the first >case in that it's a flawed test; it's cheating. > > So I guess Vincent had to be talking about professional book vs no book or >professional book versus random book, without any cooking against a certain >opponent. > > José C. Hi José If the opponent has no book, no learning and no other mechanism to randomize it's moves then it is, as you say, no test as soon as you find one winning line for white and black. I'm no book builder but my thinking would be to know the strengths and weaknesses of my engine and build a book to take advantage of the former while avoiding the latter. Then for a tournament, I don't know what book my opponent will bring, but I know that it does not handle this or that line very well (whereas I do) so I'd adjust my book to play those lines if I get the chance. I have trouble with 700 but I think such a book could easily be worth 200 or 300 elo. Best Dan H.
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.