Author: Roger D Davis
Date: 09:01:59 06/14/02
Go up one level in this thread
On June 14, 2002 at 10:48:28, Daniel Clausen wrote: >On June 14, 2002 at 10:28:24, stuart taylor wrote: > >>On June 14, 2002 at 07:42:21, Kurt Utzinger wrote: >> >>>I can only say once again that I have seen match results of 9-1 followed by 1-9 >>>and that for this reason quite a large number of games are needed to say for >>>sure that program A is stronger than program B. >>>Kurt >> >>It is a bit interesting why it goes in blocks. >>Maybe when Fisher beat Larsen and Taimanov 6-0 each, maybe Larsen too, that >>didn't mean ANYTHING in the world either? >>S.Taylor > >Ah the classic fault. > >Looking at a series of results and look for patterns is the wrong way to 'prove' >something. Afterall there has to be _some_ pattern. :) > >Little example, which is a bit extreme but shows the point: > >A: throws a die 10 times and gets 1, 3, 6, 3, 5, 3, 1, 5, 2, 2 >B: "Oh! It seems that when throwing a die 10 times, we end up with this >pattern!" >A: "Not necessarily" >B: "Yes! Yup yup!" >A: Prove it! >B: "Look at your example! Ha! I won! :)" > > >Morale of the story: >-it's ok to take an example and speculate >-you have to repeat the example (usually several times) and see whether your >claim still holds. A mathematician could even tell you how many times you have >to repeat it in order to be sure to 95% (or any other number you choose) > >HTH :) > >Sargon There is a statistical test of significance called a "One Sample Runs Test" that looks at such patterns of results and attempts to determine whether the results are not random. Interestingly, results of 9-1 would seem to indicate that the results are strongly not random, and yet, scores do seem to flip flop all over the place. Roger
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.