Author: Stefan Ottow
Date: 13:24:41 11/05/05
Go up one level in this thread
On November 05, 2005 at 11:08:46, Martin Slowik wrote: >Hi Steve, hi Stefan, > >if I may butt in... I rather agree with Stefan but would like to add just one >further point.... > >>most collectors feel this way but its true that some do not >>and of course i am greatly interested in the ratings of the high-end computers >>in my collection >> >>for the really early computers the ratings are not so important > >Actually I think most people overstate the importance of ratings way too much. >As you certainly know mathematics is what I do for a living and my experience is >that people let themselves influence by numbers too much (This statement by a >mathematician..! Well...). ;) In this particular case it's not only the >statistical problem (not enough games). The other thought is certainly not new >either: Elo numbers are just a sort of weighted averages of past results with >all the advantages and drawbacks attached to this concept. Of course if you have >two pupils, one with the average mark B the other with the mark C, you can >safely assume that the first one in general has been better. But: you still >don't know who is better, say, in math.... (Or, another example: Shirov is >certainly a strong player but the Elo numbers don't tell you the whole story >about his chances to win in the next game against Judith Polgar....). So >everything boiled down to a single number doesn't help you much - rather the >contrary is true. It gives you just the confidence to know it all.... > >Anyway. We all want to see the results expressed in one number, so let it be. I, >personally just try to keep all this in the back of my holey ole' head... :) > >>but i am only one man in a far away country and i of course cannot change what >>has already been decided on >> >>so i will go on my lonley thankless way and complain a little here and there >> >>:)) > > >*lol* > >Laughing Regards, >Martin Hi, one more thing to add to our ELO-list: Please note that we should not focus here so much on absolute numbers as Martin mentioned. With the program ELO-Stat, which we are actually using, you can generate every ELO you like, it just depends on the calibration, i. e. the starting value. We tried to find a good way to calculate and adjust it regularly, but this is just an idea and probably not the best and final solution. If you are interested you can read it here, (sorry, only in german): http://www.schachcomputer.info/html/elo_liste_-_erklarung.html Net: Its preferable to look at the differences between the machines and not at absolute numbers. calculating regards Stefan
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.