Author: Peter Berger
Date: 07:26:16 07/03/01
Go up one level in this thread
On July 03, 2001 at 09:18:14, Christoph Fieberg wrote: > >The likelihood for a comeback *after* a 0-5 defeat in the first five games is >rather high (conditioned likelihood!) > Indeed - poor wording on my side . >However the likelihhod that a whole match goes exactly like this: >5-0 lead *and* then at least equal at the end >is very low. It occurred only 55 times in 200.000 matches (0.0275%) Cool - this is in fact _very_ close to the number I expected :-)) While the result of this match was being discussed I was very much surprised about how few people seemed to realize _how_ unlikely it was in a fair match. I even remember Bob Hyatt posting that he sees results like this one "all the time" which surprised me. Yet , unlikely events _do_ happen - just not very often :) . And regarding the validity of this match there is another theory which I expect to be able to explain the results much better and without the need of "luck" . We know and it was reported during the match by one of the persons in charge that computers were swapped after game 5. Only much later it has been reported that computers were swapped more than once . Maybe this didn't happen . This is no conspiracy theory - it simply explains the results better IMHO . And in fact I am convinced _if_ this was the case that this decision was made in the honest belief that both computers were completely even and didn't give any advantage to one of the contenders. I don't care too much for the BGN match . It's interesting though that test results even when not run on a huge number of games can be checked for their validity . And : maybe it really was simply luck for Fritz after all :-). pete > >Christoph
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.