Author: Uri Blass
Date: 20:47:22 11/24/01
Go up one level in this thread
On November 24, 2001 at 17:41:21, Jesper Antonsson wrote: >On November 24, 2001 at 13:44:10, Joshua Lee wrote: > >>I know that since many problems are beyond the positional and endgame Knowledge >>of any program and even tactically programs need several hours to see the >>hardest things, is it safe to say that while most of the top programs find the >>solution to the harder proplems at about the same time that since they can be at >>16 or more hours that if the program were 540 elo stronger they would solve >>something even on the same hardware but without taking so long. >> >>For example Fritz Chess Tiger and GT2 will take 15 or 16 hours to find the >>solution to Nolot #10 on a 1Ghz using 192-256MB for HashTable. If those programs >>were strengthened 540 points or the estimated equivalent of 6 ply wouldn't they >>solve this and others in tournament time without having to have a 260+ Ghz Cpu. > >Well, there will always be positions with long range tactics that are atypical >and won't ever be solved by positional knowledge. So, IMO, some of those >positions will always need about the depth (and nodes) that they need now, and >thus a "260 Ghz CPU" (or equivalent). > >Jesper I disagree here. knowledge is also search rules and not only evaluation. I believe that the new programs are better than the old programs in solving the nolot test suite mainly because of better search rules. Another point is that the number of nodes is meaningless because programs do not mean to the same thing when they say nodes. The only thing that is important is time. I believe that correct rules of pruning illogical moves and better extensions may help program to be at least 100 times faster. The only problem is that it is not easy to find the right rules. Uri
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.