Author: Matt Frank
Date: 20:25:25 03/19/99
Go up one level in this thread
On March 19, 1999 at 16:49:09, Dann Corbit wrote: >On March 19, 1999 at 16:14:46, Matt Frank wrote: > >>On March 18, 1999 at 05:23:11, Charles Unruh wrote: >> >>>I don't think this match will really show the strength of a the computer at all. >>> This is like kasparov playing Yermo in a match. We need an average GM. It will >>>still be fun though >> >>Charles, I'm not sure that you have assessed the strength of Hiarcs 7 properly; >>but in any event the match will tell all. I assume that if Hiarcs 7 wins the >>match or draws you would then be willing to conclude that in all probability >>Hiarcs 7 deserves to be rated at GM strength, right. If Hiarcs is really only a >>2480-2510 program, Mr. Yermolinsky would be expected to score 4-2. However if >>the program is operating at 2600 + on my system, he should be no better than >>even money to hold the draw. I expect that the performance rating for Hiarcs 7, >>based on this match, will be FIDE elo 2600 +, which would imply a 3 score or >>higher for Hiarcs in this six game match. >We will need to be careful what to conclude from the match. We will be able to >state proability figures only. For example, what is the probability that a 2300 >machine could score as well? A 2900 machine? etc. > >Six games is very slight to make statistical judgements. Thirty is a magical >value that makes statistics much more reliable. > >But we will know that it is likely that it is very strong if it wins any games. I don't have my handy calculator right near, but I'll bet that a 2300 player scoring 3-3 against Mr. Yermolinsky is pretty remote (i.e., less than 5%). Yet I do agree that a safe place to be is to make probabilistic statements. Matt Frank
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.