Author: Dann Corbit
Date: 10:51:49 04/20/01
Go up one level in this thread
On April 20, 2001 at 06:16:01, Aaron Tay wrote: [snip] >I disagree. If you study games, you can develop a hypthoesis say Player X, >overvalues passed pawns or [very simple example].But it's merely a hypthoesis If >you have the program, you can see the programs scoring to help support or >dispute your hypthoesis. > >It won't be exact of course, but couldn't Kramnik with the help of some computer >experts, figure out the evulation function roughly by feeding the program >specific positions ? Of course, I'm not a programmer and I know that it's not so >simple, because evalution functions are not linear.. > >On the other hand studying Kramnik's game couldn't tell you a thing about why he >played move X.. Let's say you take one month and use 20 computers (just running at night so that they don't waste the time spent during the day). Analyze every position he has ever played for 1/2 hour per position. Store it in a database. Now, when he plays a move he has played before, you simply look at: 1. What did the opponents do? 2. What did other people do in this position? 3. What is the computer's suggested move? etc. You could also explore lines that the computer thinks would be winning by carrying them out to the end. I still think a computer could prepare for any given human with the right use of database technology. In fact, the advantage might possibly even be greater for the computer database than for the human preparation.
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.