Author: Stephen Ham
Date: 08:11:40 03/03/06
Go up one level in this thread
On March 03, 2006 at 02:58:49, Jouni Uski wrote: >More 60/5m games with native books: > > >1 Rybka Beta ½½10½10½½111½11½½½½½½111½½½1 18.5/28 >2 Fruit 2.2.1 ½½01½01½½000½00½½½½½½000½½½0 9.5/28 > >1 Rybka Beta ½11½111111½111½½1111 17.5/20 (+338 !!) >2 Shredder 9.1 UCI ½00½000000½000½½0000 2.5/20 > >Jouni Hello Jouni, How is the above a "book" test? It looks more like a blitz chess test to me. In order to test the book in the fashion that you're using, Jouni, don't you have to test the engine with its book, and then replay those matches after giving it another book, and then compare? But even then, in speed chess, how much is the result related to the book? After all, speed chess is so sloppy that I've seen many games where an engine has a very large advantage after the opening, only to subsequently lose due to one weak move, and/or a series of weak moves, played quickly due to time constraints. Similarly, I think the Hiarcs 9 book is objectively "bad". It's generally considered objectively inferior by others too. Nonetheless, Hiarcs 9 is a dominant speed chess player and its "strange" book is often an advantage for speed chess in that it takes other engines out of their books while Hiarcs remains in its book. This gains time for Hiarcs. Likewise, in speed chess, it leaves the opposing engine too little time to find what's objectively wrong with Hiarcs' book. I'm playing some matches and tournaments with Rybka and its book at very long time controls (e.g. 180/40, etc.). While the long time controls necessarily means relatively fewer games, I'm getting high quality games as a result. So far, I'm very impressed with Jeroen Noomen's Rybka book. Jouni, the above are just my thoughts at random. I'm a computer chess "dummy" so it's quite possible that what I wrote makes little sense. All the best, Steve
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.