Author: Larry Proffer
Date: 08:09:19 05/08/01
Go up one level in this thread
On May 08, 2001 at 10:49:52, Martin Schubert wrote: >On May 08, 2001 at 10:30:10, Larry Proffer wrote: > >>"Braingames explain their reasoning. "We made a simple decision. We wanted >>programs which could play on multi-processor platforms as they are obviously >>stronger candidates for the Kramnik match. There are really only three >>candidates: Fritz, Junior and Shredder. We made great efforts to persuade >>Shredder to play but they declined." They added that they unfortunately didn't >>have time for a tournament with 10 programs which would have taken too long to >>run. One of the main complainants was the company REBEL. Their TIGER program is >>a single processor prgram yet still finished second in the Cadaques event run by >>Prof. Irazoqui earlier in the year. They actually have a multi-processor version >>called DEEP TIGER but that wasn't announced until after the invitations were >>made." >> >>Can any statistician answer if it isn't actually better (in terms of finding the >>'best' comp-comp program), to increase the number of participants while playing >>the same number of games? >> >>It seems to my amateur mind that: >> >>a) a participant increase actually decreases the effect of "A beats B, and B >>beats C while C beats A" - in other words it reduces the effect of one program >>being tuned (on purpose, or just happening that way) on another. >> >>b) it decreases the effect that the 'objectively best' program, bu not actually >>playing, can't possibly 'win' the tournament. > >When you want to examine a match statistically you have to assume that these >aspects do not exist. So they don't matter. Otherwise you've got big problems to >get any statistical results. > >Martin If the objective of the match is to 'find the best program to play Kramnik', then what matters is that *all* the possible best programs are selected to compete, no? If a program that could be 'best' doesn't compete, then .... ? Isn't all the talk of "24 games", or "not enough time" not just a smokescreen? It sounds good, but means nothing? Isn't the only statistic that is actually meaningful is that the Chessbase win chances fall from 100% to 66% if Tiger plays? This is the bottom line, no? If you add Crafty and Ferret to the mix, Chessbase win chances fall to 40%. I don't see any other meaningful statistic. Number of games is irrelevant. Therefore time needed is irrelevent. They used five days, they could have done any other combination of programs in that five days. The 'validity' of the winner being 'best' is not affected by time or games. The 'validity' of the winner being 'best' is affected only by the pool of programs used. Exclude a program, or exclude three programs, and the 'validity of best' crashes in inverse proportion to the 'Chessbase win chances'. That's all there is to it.
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.