Author: Micheal Cummings
Date: 17:52:26 12/05/98
Go up one level in this thread
On December 05, 1998 at 10:45:30, Nelson Hernandez wrote: >Michael, I tried your settings vs. out of the box CM6K for a 40-match >tournament, 5 minutes blitz. Your settings narrowly prevailed, 15-13-12. > >My observation is that it was really dead even in most matches after 60 moves or >so. Your settings came out on top because in a few of the longer games (90+ >moves) CM6K's standard settings were more prone to make a critical error late in >the endgame when it only had time to analyze to the first ply. > >My conclusion: five minute blitz isn't really a fair test. What do you suggest >might be a better one? My settings I think work best over 40/2 hrs. My settings were not created by testing blitz games, I should have made it clearer. I only said at the end of the post that I did play a few games of blitz, but this was not the test for these settings. I only played a limited amount of 5 minute games, but in the ones I did play they won clearly, I run my games on two 450Mhz machines with 128 ram. But my settings are for 40/2hr game controls. I should also say that you should increase the attack of CM6K by 6 points in Blitz games than the ones I posted. But these are not blitz settings. Sorry I should have made that clearer. Plus if you played the games on one machine, then I do not think the games show really the true result. I did a few games on one machine and found the program did not perform at its best. even with deep thinking turned off. CM6K just hogs to much CPU time, even if playing personailty against personality. I think you need two seperate equal machines to get accurate results Plus I should also say that my settings include a 64 meg hash, not the standard one meg hash.
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.