Author: Enrique Irazoqui
Date: 04:44:42 09/14/00
Go up one level in this thread
On September 14, 2000 at 01:46:31, Dann Corbit wrote: >On September 13, 2000 at 23:29:49, Bruce Moreland wrote: > >>Something I find confusing about this group is that there are a zillion >>tournaments going on at all times, and if you look at the post titles, it's hard >>to know if they are tournaments involving computers competing at some site with >>their authors consent, human tournaments, or championships of someone's >>basement. >> >>As far as I can tell, this is the championship of your basement, right? > >It's the *OFFICIAL* world championship of my basement. :):):) > Actually, I think it is >far more relevant than any of those ICCA thingies. They run so few games that >the winner is almost a pure crap-shoot. Well, maybe not necessarily crap, but it may happen easily. I think we can trust the Elo list of FIDE, which counts hundreds of games of each player, much better than the "world championship" of Las Vegas. A ranking of 1.Kasparov, 2.Anand, 3.Kramnik, 4.Shirov, etc., makes more sense to me than 1.Khalifman, 2. Nisipeanu, etc. For the same reason, I trust the SSDF list much better than the WCCC. Junior 4.6 won a WCCC, but I would bet for Rebel 8 any time if both had to play in the same long tournament. Wouldn't you? Same for Shredder 4 and Fritz 6. A 7 game event is just not enough, not even close by a long shot. I guess this must be why a programmer we all know feels more proud when he tops the basement SSDF list than after winning the loft WCCC, and he did both. I would feel the same way. Enrique > By running at G/60 and playing 4 games >for each round robin pairing, we will achieve _actual data_ with _actual >relevance_ as to the approximate strength of the programs. Sort of like SSDF, >but the error bars will be considerably larger. By the time the contest has >completed, there will be about 650 games played between the programs in the 4 >crown division, and well over 1000 games overall. > >My championship will not carry with it the prestige of the ICCA title, but the >winner (and even more so the programs that did not win) will know far more about >the strength of their programs than the data received from any contest with a >tiny smattering of games by each program, often with only one color against a >given opponent [or worse yet, with Swiss format affairs -- quite a few opponents >not even attempted]. > >I usually have 4 or 5 machines going every night, with 4 games apiece as >mediated by Winboard. The machines have to be PII 300 MHz or faster. > >Each program gets 32 megs of hash to work with.
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.