Author: Robert Henry Durrett
Date: 10:46:18 09/05/98
Go up one level in this thread
On September 05, 1998 at 12:51:26, blass uri wrote: <snip> >I believe that the programmers can know better which version is better because >they have more information(they know the source code of their program when you >can see only games) > >The question is if the reason that tiger11.2 won is the opening or another >reason(you or christophe can check it by playing the game with the same opening >with opposite colours) > >Uri Perhaps one benefit of comparing a recent version to an earlier version is that, if the more recent version does poorly in comparison, then the presence of serious unintentional new bugs or programming errors in the new version will be detected. Also, if the new version is better, running enough test games might answer the question "How much better?" [Within an acceptable level of confidence]. But, and this is a question, is not the primary purpose of these tests to see if the intended improvements have been made? If so, then wouldn't running tests using test positions be better for that purpose? It is not clear why the two or three above games are being run, except for "general interest." How many games are planned?
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.