Author: Robert Hyatt
Date: 17:38:21 05/12/01
Go up one level in this thread
On May 12, 2001 at 10:54:42, Joshua Lee wrote: >>>The fact that deep blue played a different move does not prove nothing. >>>Both Rxc6 and bxc6 are losing moves and different extensions can encourage a >>>program to play a different move. >>> >>>We do not have the logfile of deep blue from these games so we can know nothing >>>about the line that Deep blue saw and if it changed it's mind from Rxc6 to bxc6 >>>so it seem that using this position to compare between the first Deep Blue and >>>the commercial programs is useless. >>> >>>Uri >> >> >>That's why I asked. First, DB1 had several shortcomings that were addressed >>in the new chip used in DB2. Second we had no output from DB1 as you said, >>so there is no way to know what it saw and what the eval was (of course, for >>DB2 it is not easy to know what it saw either since it can not display all >>(or even most) of the PVs. > > >This Comes close to giving me an answer, I don't understand how exactly Deep >Blue 1 goes for the so called worse move but Fritz goes for the other even if >you let it think to the same amount of nodes. If there were bugs in both Deep >Blue 1, and Deep Thought how is it possible to compare comercial software? This >i believe is needed as well as a variable Eval to furthur improve. My other >guess at comparing the programs to Deep Thought or Blue is to find a point in >several different games where it made the mistake and see how long it takes >Crafty, Hiarcs, Fritz etc. I think this would be interesting to know however >some positions are way too deep for my single 800mhz i am beginning to think i >need a beowolf :) The first test is to let fritz search both moves to the same depth to compare the scores. It might be that they are _very_ close after a bit...
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.