Author: Dann Corbit
Date: 16:07:04 12/29/99
Go up one level in this thread
On December 29, 1999 at 18:57:36, Michael Cummings wrote: [snip] >Lets take the world cup soccer for example. I do not really follow soccer but I >am sure that is one sport that everyone would know about. France won and they >claim themselves to be the best in the world, Just like Shredder winning the >WCCC99. Now in soccer, brazil is still ranked number one. So by the the world >cup rules France are No1, when in fact most would claim that they maybe around 4 >or 5 from what I hear. > >Of course results are valid under the given terms and conditions, does not make >them accurate in the logic of things. > >If you take the SSDF list as it is now a true reflection of which is the >strongest program, you are also proving my point that there is no way shredder, >which won WCCC99 is the best in the world, which goes to prove my point that the >winner is not always the best. I don't think it proves anything of the sort. But I think that the SSDF is a much better indication of playing strength than any single, short contest. Consider a one game soccer match. If both teams are fairly evenly matched, then it's approximately a coin toss who will win. But if one team is 70% stronger, the other team still has a good chance to win one match. But if they played 30 games, the "truly best" would be much more certain. However, even after one million games, which was the best would still have a tiny, statistical doubt. >The plain fact is Shredder won the WCCC99, and Tiger is on top of the Chess >rankings on SSDF, so without going into anything else, which is the >true number 1 ???? You are correct that there is no answer to this question [the question being "Which program is stronger?"] >The chess rankings clearly prove what I am trying to say. You make a very good point that I think most people do not understand. Being the top program in a contest or an SSDF list is not a formal proof of the highest strength. It is only an indication. The more data we gather, the stronger the indication. But it can never be proved in this way, unless the number of games approaches infinity. However, it is still interesting to perform the experiments because we do get statistical indications. And if a program is a full 3 standard deviations above a different one, we could say with very good certainty that it was stronger. But when programs are all overlapping withing a single standard deviation (as are the top programs in the SSDF) we really don't know for sure which is stronger. It's nothing more than a 'best guess' with mathematical basis.
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.