Author: Charles Roberson
Date: 15:12:01 02/17/05
Go up one level in this thread
Some experiments I did a few years ago with NoonianChess revealed that
a good book can make a difference.
I ran a guantlet with NoonianChess using a particular opening book.
Then did it again with a different book.
There were opponents in the first experiment that NC lost 0-4 against, but
in the second experiment the results were reversed. Also, there were
opponents that NC faired better with using the first book. In particular
NoonianChess was regularly shut out (0% score) against Gerbil with book #1.
But, NC won 100% of the matches against Gerbil using book #2.
Calculating the differences in results if a combined book were used and
openings targeted to specific opponents revealed around 200 elo point gain
in rating. This proves that a 200 pt rating improvement may be achieved
in some cases.
The first book was made of my favorite openings and the second was the
CAP data from Dann Corbit. Neither was especially tuned for any opponent.
So, I could easily believe very well tuned books could be worth as much
as 400 elo points maybe.
Charles
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.