Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rating in ICC is meaningless and here is an example

Author: Uri Blass

Date: 08:28:25 01/14/03

Go up one level in this thread


On January 14, 2003 at 11:14:09, pavel wrote:

>On January 14, 2003 at 11:02:30, Uri Blass wrote:
>
>
>
>***** RATINGS *****
>
>The Basics:
>-----------
>You can have many different ratings on the ICC.  There are ratings for bullet,
>blitz, standard, 1-minute, and 5-minute, which are different speeds of
>regular chess.  There are ratings for bughouse, loser's, and wild, which
>are for chess variants.  See "help definitions" for an explanation of
>what these mean.  Type "finger" to see your current ratings.
>
>To get a rating in one of these categories, you only need to play
>a rated game.  Do "set rated on" and then do "seek" to ask for a game.
>You can ask for any time control or chess variant you wish.  Type
>"help seek" for more information.  Unrated games will not count for
>your rating, but you are welcome to play unrated games.
>
>Ratings usually range from 600 to 2800 on ICC, but there is theoretically
>no limit at either end.  Beginners usually get ratings from 800 to 1200.
>People with ratings over 2200 are considered to be "masters".  Grandmasters
>playing on ICC usually have ratings over 2400.  You can type "best"
>to see the highest-rated players on ICC.  You can type "rank" to see where
>you stand among ICC members.
>
>Ratings on ICC are similar to the USCF and FIDE, but are totally separate.
>Do "help survey" for a statistical comparison.
>
>A player's rating is "provisional" if he/she has played less than 20 games.
>A rating is "established" if it is based on 20 or more games.  A different
>formula is used to calculate ratings for established and provisional players.
>See "help provisional" for some information about provisional ratings.
>
>Everything you wanted to know about rating formulas, but were afraid to ask:
>----------------------------------------------------------------------------
>The rating during the provisional period is the average of a set of values,
>one for each game played.  The value for a game against an established player
>is the opponent's rating plus 400 for a win and minus 400 for a loss.  For a
>game against another provisional player, the value is moved towards the
>previous average to lessen the impact of the unreliable result.
>Extra points are then added to the rating for the purpose of keeping the
>average rating of all established active players close to 1600.  In particular,
>1/5th of 1600 minus the current average is added to the rating.
>To explain the established period requires the use of a formula.  Suppose your
>rating is r1, and the opponent's is r2.  Let w be 1 if you win, .5 if you
>draw, and 0 if you lose.  After a game, your new rating will be:
>
>                                      1
>              r1 + K * [ w -  ---------------------- ]
>                               1 + 10 ^ ((r2-r1)/400)
>
>I still need to explain the variable K.  This is the largest change your
>rating can experience as a result of the game.  The value K=32 is always used
>for established player versus established player.  If you're playing a
>provisional player, the factor K is scaled by n/21, where n is one plus the
>number of games your opponent has played.  See also "help k-factor".
>
>This formula has the property that if both players are established then the
>sum of the rating changes is zero.  It turns out that if the rating difference
>is more than 719 points, then if the strong player wins, there is no change in
>either rating.
>
>Note that during the provisional period, BEATING a player whose rating is more
>than 400 points below yours will DECREASE your rating.  This is a consequence
>of the averaging process.  It's useful too, because it prevents the technique
>of getting an inflated provisional rating after one game, and then beating 19
>weak players to get an established rating that is too high.


Sorry but a technique that is useful to prevent getting inflated rating and does
not prevent getting rating that is too low is not useful.

I do not see what is the problem to change it and decide that if the winner lost
rating the game is not counted or even better to prevent players to play against
players that they can only lose rating against them.

It is clearly a bug or a design flaw in the program that calculates rating .

Uri



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.