Author: Robert Hyatt
Date: 06:38:42 08/09/01
Go up one level in this thread
On August 09, 2001 at 07:15:36, Uri Blass wrote: >On August 07, 2001 at 23:10:18, Robert Hyatt wrote: > >>On August 07, 2001 at 16:03:34, Bruce Moreland wrote: >> >>>I've often heard people state that null move with R=3 is better than with R=2, >>>but I have never ever ever gotten a test result that indicates this. >>> >>>I've tried everything. I've tried it throughout the tree, I've tried it near >>>the root, and I've tried it near the tips. >>> >>>My measurement standard is ECM positions solved, which *always* goes down. >>> >>>What are other people doing that I'm not doing, or are people testing in some >>>other way, if so is their way better or worse? >>> >>>I would test Crafty both ways (it's currently doing R=3 some places), but my >>>machines will be busy until after the WMCCC. >>> >>>bruce >> >> >>I don't think it helps on test suites at all. But it does make normal searches >>significantly faster, which is why I use it... > >I believe that people simply use the wrong test suites. > >Test suites should be based on games positions and not on tactical positions and >it is possible for example to take all the position of part of the games of last >WCCC as a test suite. > >The only problem is to decide about the solution of the test suites and the way >that I suggest is to take some top programs and to give them to analyze for many >hours every position. > >After doing it you can get one move or more than one move for every position. >The set of moves that you get for every position can be defined as the solutions >of it. > >It is possible to remove part of the moves from the set of solution in cases >that humans are sure or almost sure that the moves are not correct solutions >inspite of the fact that some programs liked them after many hours. > >Not all the solutions are going to be correct but I expect programs to solve >more positions when they get more time and I expect most of the positions not to >be tactical. > >Uri This is pretty hard to do. If a position is supposed to be a mate in 5, that can be confirmed pretty easily. If it is supposed to have one best positional move, that is more complicated to prove. As it is always possible that a very strong tactical program will still find a tactical answer to the position, and then trying to confirm this will take time. It would be a nice test of course, but it would probably go through continual revisions because positional answers are less concrete than direct tactical wins.
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.