Author: Thorsten Czub
Date: 11:53:35 09/08/98
Go up one level in this thread
On September 07, 1998 at 03:25:59, Ernst A. Heinz wrote: >On September 06, 1998 at 14:07:42, Thorsten Czub wrote: >> >>SELF test ?! >> >>I don't think SELF test (x vs. x+1) makes much sense. Sorry. > >Thorsten, > >Although you might never have thought it, I mostly agree with your disdain >of self-play. It is always a compliment to me, if people with different background do agree with me :-)) So - thanks. >It seems to be a necessary evil like test suites (just to be >sure you did not break the new version) but it does not tell you very much >about the real improvement in playing strength -- especially not if the >changes of the new version are mostly added or tuned evaluation features. >In case of pure search enhancements (e.g. new extension or pruning methods), >however, self-play is quite reasonable and expressive. Thats right. If i want to find out about a special feature of the search, i have changed, thze data is important. Of course. I only want to warn people to be to shallow and ONLY use x vs. x+1 as the base of their work. >=Ernst=
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.