Author: Michael Cummings
Date: 16:06:12 12/29/99
Go up one level in this thread
On December 29, 1999 at 13:53:03, Dann Corbit wrote: >How to measure what "best" means is difficult "at best." > >Consider human players. Is the world champion the best or is Kasparov? >Is Anand better than Shirov? > >I think (probably and only in my opinion) that Kasparov is the best player in >the world. But I could be wrong. And we have a very large number of games >available to make a decision. > >Now, the SSDF values hold for exactly: >0. The machines used >1. The programs used with that particular time control and settings >2. The method of play used > >And we still have a large uncertainty. > >The WMCCC champion is a proclaimed champion, like A. Khalifman of FIDE who may >or may not be the best player in the world. and B. Gulko of USCF who may or may >not be the best player in the US. In any case, we know both of them are darned >good players. > >Test suites like BS2830 try to measure ELO and do manage to measure some form of >tactical prowess. But how neatly does this map to playing strength? Really, >nobody knows. > >Another measure of strength is the rankings on ICC or FICS. > >Yet another is the results from Winboard tournaments like those defined at: >http://f11.parsimony.net/forum16635/ > >Quite frankly, all the programs are very, very strong. The tactical ability is >probably GM strength [maybe even super-GM] and the only real weakness is long >term strategy. > >Since all of them will beat the pants off of anybody but an IM or better (and >still give IM's and GM's a hard time) strength should probably be the least of >our worries. Really, we might be better served by looking at the sort of >analysis features that are offered and database capabilities. > >In short, the best program is the one that is best for you. > >As far as strongest, nobody knows -- really. For computer verses computer play, >the SSDF list is probably the finest tool. But if you don't understand some >statistics, it would be foolish to try to draw conclusions from it. In >particular, any of the top ten programs could easily be better than the others >by simple examination of a single standard deviation. Correct observations, you can claim that a program is number 1 and state that because the SSDF said so. But in reality as you explained this is not totally the correct result. As I said in a post below, WCCC99 Shredder or SSDF Tiger, which is number 1 right now given the results.
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.