Author: Robert Hyatt
Date: 11:02:27 09/18/05
Go up one level in this thread
On September 18, 2005 at 13:40:28, J. Wesley Cleveland wrote: >On September 18, 2005 at 11:17:42, Robert Hyatt wrote: > >>Note that this "experiment" will be a tough one to run. One could run a match >>1000 secs/move vs 1 sec/move. I suspect the null-move crafty gets killed there >>as the searches will be shallow enough the null-move failures will significantly >>influence the games. 1000 minutes (about 1 day) vs 1 minute will make a game >>take a couple of months. Also not good for a large experiment... > >There are significent differences between between matches played on different >speed hardware and matches at time odds in the areas of pondening and time >management, e.g. in the experiment above, the null-move crafty will outsearch >the other crafty by several ply each time it ponders the right move, and will be >outsearched by several ply otherwise. can't do pondering in time-odds matches, when the primary goal is to pretend one of the programs just has much faster hardware in an equal time-control match... ponder=off is the best approximation possible...
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.