Author: Sune Fischer
Date: 07:30:38 06/18/04
Go up one level in this thread
On June 18, 2004 at 10:16:55, Sune Fischer wrote: >On June 18, 2004 at 09:47:55, Rolf Tueschen wrote: > >>Thanks for the information. Let me add in return something important which would >>also explain the big interest of the journal CSS. >> >>Mikhail G. sees himself at eye-height with the programmers of the chess >>programs. He claims that he has worked on the analyses of "his" positions for >>years! And he sees himself justified because the test results from Manfred >>Meiler do show that the CSS-WM-Test ranking list (for actually some 300 engines) >>is similar to those lists that are made to test playing strength. So, Mikhail >>claims that his test is ALSO a good indicator for the playing strength, and >>more, in case of criticism, Mikhail can make the counter argument that his test >>does only deals with "analytical abilities" which is allegedly the base of all >>strengths. So, this is the real reason for the mass hypnosis of the dear CSS >>readers. So, even those who dont like MG still can have the delusion that >>"their" test is kosher because there is no critic against it. > >For this to be true, it would have to be a measure of strength equivalent to >playing games, IMO (at least ignoring books, time management and stuff). > >Imagine if everyone starts developing for years only to score high on that >suite. >Do you think the one scoring the highest (assuming they don't all get 100%) will >also be the strongest? > >It's quite obvious it's not that simple in general, just compare speeds of CM >and Fritz in finding mates and see how that relates to their chess strength. > >To make a balanced suite where these two engines end up close to each other >you'd have to find a whole different set of positions where Fritz scores better. > >Now you take a third engine, say Junior, which scores badly on the Fritz suite >and the CM suite because it has its own set of strong and weak points. So you >have to add a new suite for Junior. Etc. for a fourth engine.... > >This kind of reverse engineer patching of a testsuite will by definition produce >the right strength relations, but still it does not indicate strength. Actually I want to correct that, I think it might more or less vaguely "indicate" strength, but it does not "measure" strength, and that's a big difference. -S. >-S.
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.