Author: Robert Hyatt
Date: 14:35:43 01/31/99
Go up one level in this thread
On January 30, 1999 at 11:22:59, KarinsDad wrote: >On January 30, 1999 at 06:33:21, blass uri wrote: > >[snip] > >>>My meaning is as follows: >>> >>>Win expectency for a difference of 1000 points is 0.00315231 (as per Dann's >>>posting). >>> >>>This means that out of 100,000 players with an 1800 ELO (1900 USCF rated >>>players), 315 of them should on average win (i.e. 315 wins or even more draws) >>>against Garry Kasparov in standard tournament times if all 100,000 of them >>>played him. This is total bull. It would be an amazing event if even one of them >>>won (or even drew) against Garry Kasparov in standard tournament times. >> >>I do not think it will be an amazing result. >> >>Kasparov is a human and can do a stupid mistake if he play many games. >>If kasparov is suddenly ill in a game and the opponent did a good opening >>preperation against him then kasparov may lose. >> >>I know a case when a 2000 elo player won a grandmaster in a tournament game >>after less then 2 hours. >>I asked him how did he do it and he replied that everything was opening >>preperation. >>The grandmaster fell into a trap(The 2000 player prepared at home) > >What level GM 2300, 2400, 2500, 2600? In any of these cases, the formula >probabilities are: 300 3 in 20, 400 1 in 11, 500 1 in 20, 600 3 in 100. > >These statistical ranges may be close to accurate, but I doubt it. But you'll >notice that all of these ranges are well within what we would expect to see in >our lifetimes. It should be easy to see a 2000 ELO player beat a 2600 ELO player >if it is 1 chance in 33. Sooner or later, you will see it. Maybe 10 times a year >in the US. Even if it was 1 in 300, you would eventually expect to see it >happen. Probably about once a year in the US. So, your example is anecdotal at >best. You expect it to happen eventually. > >So instead, to illustrate my point, I was talking about a 1000 difference where >the PRACTICAL chances are almost statistically 0, but the formula states that it >is 1 in 317. Yes Kasparov could get sick and lose a game or two. But he would >not lose 315 games out of 100,000 or 3 games out of 1000. > >I personally think that Kasparov could be on this deathbed and play 1000 games >against 1000 1800 ELO players over the last year and a half of his life and win >them all, but that is pure speculation. > >Again, it would be nice to analyze results of all of the rated games in the US >for the last year. That would be a sample size of what, between 200,000 and >500,000 games? Give or take. That is what we need to find out if I am correct or >not. > >KarinsDad All of the Elo analysis is based on the 'normal curve'... which absolutely says that a 2600 player can lose a 10 game match 10-0 to a 1000 player. The odds aren't very good, to be sure, but the probability is not 'zero'... which is probably the point of the whole discussion?
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.