Author: Amir Ban
Date: 13:29:27 09/01/98
Go up one level in this thread
On September 01, 1998 at 13:54:54, Mark Young wrote: >On September 01, 1998 at 10:19:05, Amir Ban wrote: > >>On August 31, 1998 at 17:57:02, Mark Young wrote: >> >>>Is computer Vs computer testing now useless in gauging a chess program’s >>>strength playing humans? When Crafty gets killed playing Junior 5 by a wide >>>margin. And Fritz 5 draws a match with Rebel 10 even when Rebel 10 has a 2x >>>hardware advantage. >> >>Sounds like you are saying that comp-comp testing show such great rating >>differences that they can't possibly be realistic. > >But the results are skewed. Even in the SSDF testing if one program can get >one-sided results against one or two programs it will suggest a bigger advantage >against humans, Because it is #1 on the list. And with only a handful of >programs being near the top of the list there is not the pool size to correct >the over inflated rating. I am not saying that fast searches can not be the best >at playing humans. I find fritz 5 to be the best program I have tested so far >playing humans. But if computer Vs computer testing squashes other programs that >may be better at playing humans or just depresses the rating so much that the >general public will pass by the other programs that would also play very well >against humans seems very counter productive to the computer chess community as >a whole. And I only say this because the reason for computer Vs computer testing >was to give the public an idea of which programs are stronger and which ones are >weaker in general overall strength. Not which programs are best playing each >other but excluding humans. I don’t think this an issue of fast searches Vs slow >searches when it comes to playing humans. I think it has been shown that either >concept can play very well against strong humans. > > I didn't mention at all the issue of fast-searcher vs. slow-searcher, so I don't know why it comes up in your reply. The point I was making is that, for all we know, comp-comp testing also reflects comp-human results, and perhaps rating produced this way are really "objective". I'm not saying that this is necessarily true. I'm saying that this is the default assumption, and that the evidence we have points this way. People who claim that there are programs who will rate lower when measured in comp-comp games but are actually stronger in comp-human competition had better come up with something to substantiate that. Otherwise why should anyone assume that this is true ? Amir
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.