Author: Don Dailey
Date: 09:01:01 01/18/98
Go up one level in this thread
My program is similar except we are very stingy about extensions. I am running on a single Alpha for this test so we could adjust down for hardware but I don't think that is necessary. I think the fast micros are still outsearching us by quite a bit because our evaluation is 2/3 of our time. It seems like Nimzo was doing 250K NPS on a 300 MHZ pentium and I'm only doing 125-175K NPS in middlegames on 1 processor at 466 MHZ. I don't know if Nimzo is typical or not for the good micro's. The question is should we scale down a little for hardware or just use the numbers I have now? I am still running the set but only have 100 problems left or so. I ran them about 30 seconds to get an idea of how they are distributed in difficulty. I'll email the problems to you I cannot solve in 10 seconds so we can "or" them together. When we come up the final set I will volunteer to do a pass which might eliminate multiple solutions or other problems. After I do this I'll forward the quirky problems (if any) with information for each problem on why they should be eliminated and we can check each others work this way. Ed has volunteered a repository. Do you want to be the editor and keep the set up to date? I can do this if you're not inclined. - Don On January 18, 1998 at 11:00:27, Robert Hyatt wrote: >On January 17, 1998 at 23:57:02, Don Dailey wrote: > >>Ok I'm running the problems tonight to cull the easy ones. The ones >>my program cannot find in under 10 seconds must remain in the set >>assuming they are not bad problems as we discussed. When we have >>a final set I/we can test for multiple solutions. I can hack my >>program to not include they key move in the root move list and look for >>multiple solutions (near the same score.) Who is the third volunteer? >> >> >>- Don >> >> >> > >one important point. Everyone knows what sort of "searcher" Crafty >is... >full width for N plies, several extensions, simple q-search. How would >you classify Cilkchess? And we need a third volunteer that has a >different >sort of search, like maybe Genius, or Rebel, or something that has a >different approach... So that we don't stack the deck hard for our type >of searches, but produce something easy for other (perhaps selective) >searches...
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.