Author: Thorsten Czub
Date: 16:25:11 01/25/00
Go up one level in this thread
On January 25, 2000 at 17:34:50, Enrique Irazoqui wrote: >>junior kills shredder. > >Not really. 100 Elo = 6.25-3.75 >>suddenly the results become worthless. >>like letting kasparov playing a 1500 ELO guy. >>of course the 20 games match between kasparov produces measurable >>data, as much as the car-crash produces measurable data. >>but what both things connects is: >Nonsense. If Junior 6 is 100 points better than Shredder 2, both on the AMD450, >then Junior still on the 450 should be about 170 or 180 better than Shredder 2 >on the P200MMX. Say more or less 7-3 instead of 6.25-3.75. That's what you don't >understand. enrique. its not a question of maths. its not a question of the elo-formula. its not a question of understanding. when i do a poll with people, i do not only ask the ones with the blind-sign on the arms, because this would influence the outcome of the result. when i test chess programs, i try to test the programs within the ranges they are used to be. when i have chess programs, i do not let them swim in my bathtup to see if they can swim! and i do not test the one machine 2 times faster than another, against a 2 version older program. of course i can let them swim and watch out which cd sinks the fastest. of course i do get results. of course i can put these results into a formula. of course of course. but only of course. you can do it with humans. because humans do not play 2 times faster. or IF they do suddenly play to times faster, than they are doped. and therefore you forbid doping. because it would influence the data. you see: it influences the data. of course you get results. of course the data makes sense. but the question is not if the maths works, or if the formula works. the question is: do you test what you want to test ? so when you want to test the strength and test how they would swim in your bath-tub, you would not test the strength. imo you would not find out about juniors strength. understand: it is not maths. it is not arpad's formula. it is you making the mistake by only asking people having a blind-sign on their arms. >One more point you forget: a program gets a rating when running on a given >platform and you can't really dissociate one from the other. If you suddenly use >only new platforms for all programs, all references disappear and with them all >ratings. it needs to be measured again. cause the quality of the range has changed. when you relate e.g. a trabbi with a todays car, you cannot really relate them. i mean: you can try, but you will find no really good value, because you cannot measure different qualities of cars that lived in different times. you have to standardize it again and again. example: the IQ test have the same incest problems. not because the people get really more intelligent. in fact they get more stupid. but they get higher IQ's BECAUSE of this incest problematic i tried to talk with you. or in other words: of course the cars get better, but in fact the difference is not that big as it looks like. since you cannot measure this accurate without getting this range problem. try to understand it with the elo or with the IQ example. The elo rates increase and increase although the people are not really stronger. this IS an effect of the incest problem. and there is only ONE reason: you have to cut.- and to make a new beginning with new standardisation. because otherwise you get inflated IQ's and inflated ELO's. thats not a question of maths !! and not of understanding. thats a question of methodics. >This is a not-sensible comparison. It's 170 points difference we are talking >about, and not 1300. In other words, you are saying that Kasparov should never >play an under 2680 Elo player. Nonsense, isn't it? he can play against them. but the elo-difference BOTH get will not be relevant. its like testing something that is not what you want to test. when doing an IQ test with goethe and with kasparov, what do you think your test will find out ? I tell you: kasparov will get a higher IQ. this is: Junior gets a higher ELO this is: CD sinks slower in the bath-tub than hd-disk but do you really believe that kasparov is more intelligent than goethe or leonardo da vinci was ? thats the difference between quality and quantity. you try to measure quality via quantity. this works when BOTH objects live in nearly the same range. if they come from different ranges, so out of the range, you need to standardize your methods again. otherwise you will come to strange results like: the chessplayers get stronger and stronger (says your elo = junior is better than shredder2) but in fact this is not true. >Thorsten: this is the danger of thinking in metaphores: feathers and cars. It >doesn't work, it never did, it can not, it's bad thinking. pah-. what works with feathers and cars and IQ or ELO is working with in any other field too. anybody knows this !! nobody believes that the people really get more intelligent, just because your IQ tests show that. anybody knows that your test is not anymore able to produce sensible data. so they switch from 200 Mhz machines to 450 machines and begin from sero again. only in computerchess and Elo-system they produce incest, incest, incest. and the elo ratings increase, DUE TO THE WRONG METHODS. >Sorry for preaching. >Enrique quality and quantity.
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.