Author: Robert Hyatt
Date: 07:27:01 02/01/99
Go up one level in this thread
On February 01, 1999 at 00:39:27, KarinsDad wrote: >On January 31, 1999 at 17:35:43, Robert Hyatt wrote: > >[snip] >> >> >>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? > >Of course it is never zero. My point in the discussion is that the curve is too >broad (i.e. the percentage chances are too high at the extremes). > >KarinsDad I believe that the ratings near the extremes are 'compressed'. Because in the middle you can play folks above and below you, so that if you are better than the 'group' you are in, you can move up. But if you are at the top, you only separate yourself from those lower than you by a max that is very hard to overcome. IE take a bunch of 1400 players and one GM. If you start off with the 1400 players rated, and toss the GM in, the GM will quickly rise to 1700-1800, but to go further is _very_ difficult, since there is no one close to his rating he can play... That's why I usually say I think the Elo scale is 'exponential' near the ends because the actual skill difference between a 1400 and a 1600 player can be much less than the difference between a 2700 and 2800 player... Of course if the 2700 and 2800 players play _each other_ a lot, that will tend to separate them by the 'proper' amount...
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.