Author: Moritz Berger
Date: 02:24:29 10/04/98
Go up one level in this thread
On October 03, 1998 at 17:14:21, Enrique Irazoqui wrote: >It doesn't matter to whom? It matters to Blass, since he has an unlearned >Powerbook. It matters in fact to everyone. What you say above is true in theory >and doesn't apply in practice. None of us has the learned Powerbook after 2,000 >games. It doesn't exist. You could at least take all publically available 40/120 games. I guess you would end up with several hundred. >I agree there is this huge difference you mention between a learned and an >unlearned book. And, if I am not mistaken, this is precisely Blass' point. > >I think it matters to know what is being measured when we quantify and rate. Don't forget that a book with optimal weightings against other computers is very likely not the best choice against human opposition. I'm mostly interested in how a program performs against humans, no anti-computer-killer book can help here. In fact, since every player has his individual preferences, *relative* performance even in openings that "computers don't understand" could be great vs. humans if the computer plays into its opponents weaknesses. In my opinion, the PowerBook is a good starting point. If you want to maximize performance, you have to do some adjustments via learning or manual editing of book weightings. Moritz
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.