Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old July 19th 05, 11:13 PM
JP Hendriks
 
Posts: n/a
Default What about this rating calculation?

Suppose player A has a rating of 1700, player B has a 2000 rating.
They play a 100 match game which ends in 50-50.
This is obviously a great result for A and A can be considered about
equally strong as B. But wat happens to the ratings.
According to the FIDE expectation table A was supposed to score 15
points. Assume a K factor of 20 then his new rating is:
Rnew = Rold + K(Realized-Expected)=1700+20(50-15) = 1700 + 700 =
2400
For B the rating result is:
Rnew = Rold + K(Realized-Expected)=2000+20(50-85) = 2000-700 = 1300

This is definitely not a fair result. A now playes at grandmasters
level, and B has become a beginner.

Is there something wrong in the rating calculation; is this a "bug" in
the formula's. Is this problem covered anywhere. I love to get the
option of some rating experts here.

Thank you,
JP.

  #2   Report Post  
Old July 20th 05, 12:14 AM
bruno
 
Posts: n/a
Default

In Belgium there is a rule that if in 1 period (6months here) you played
more than 25 games, and your rating varies with more than 150 points then
your new rating will be equal to your performance. So in this case the new
rating of player A would be 2000 and the new rating of player B would be
1700 which is still not 100% fair it seems but ok.

I don't think this rule is used by fide though, what I have heard is that
there used to be a rule that your rating couldn't change more than 100
points during one rating period but that they abandonned this rule because
it had little purpose. A 100 game match in a period of 3 months just isn't
going to happen and would never be accepted for calculation by fide anyway.


"JP Hendriks" schreef in bericht
...
Suppose player A has a rating of 1700, player B has a 2000 rating.
They play a 100 match game which ends in 50-50.
This is obviously a great result for A and A can be considered about
equally strong as B. But wat happens to the ratings.
According to the FIDE expectation table A was supposed to score 15
points. Assume a K factor of 20 then his new rating is:
Rnew = Rold + K(Realized-Expected)=1700+20(50-15) = 1700 + 700 =
2400
For B the rating result is:
Rnew = Rold + K(Realized-Expected)=2000+20(50-85) = 2000-700 = 1300

This is definitely not a fair result. A now playes at grandmasters
level, and B has become a beginner.

Is there something wrong in the rating calculation; is this a "bug" in
the formula's. Is this problem covered anywhere. I love to get the
option of some rating experts here.

Thank you,
JP.



  #3   Report Post  
Old July 20th 05, 09:58 AM
David Richerby
 
Posts: n/a
Default

JP Hendriks wrote:
Suppose player A has a rating of 1700, player B has a 2000 rating.
They play a 100 match game which ends in 50-50.


This is such an unlikely scenario that the rating system has not been
designed to cope with it. Remember that FIDE Rating periods are only
three months long so it's very difficult to play 100 rated games within a
single period.

Also, your assumed K=20 is a little out -- FIDE uses K=15 for established
players (those who have played more than 30 games) and K=10 for estab-
lished players whose rating has ever been over 2400. Even K=15 will give
the unfair results you describe for this match, of course.


Dave.

--
David Richerby Salted.com (TM): it's like an
www.chiark.greenend.org.uk/~davidr/ E-commerce portal but it's covered
in salt!
  #4   Report Post  
Old July 25th 05, 03:02 AM
JP Hendriks
 
Posts: n/a
Default

David and others,

I agree that this is not likely to happen. But I face a problem of calculating
club ratings for an annual event in the sense that it actually lasts 30 or 40
playing rounds. Players with a too high start rating will lose too many rating
points, because of their heavy distance of estimated and realised score. Next
season, for the same reason their rating is too low, and win much more than
realistic. They become jojo players.

I would recommend rating adjustment after each playing round for long tournaments.
The extreme case I gave before results in better ratings, and I assume this will
also happen in my 40 rounds tournament. Does anyone see serious drawbacks?

JP
David Richerby wrote:

JP Hendriks wrote:
Suppose player A has a rating of 1700, player B has a 2000 rating.
They play a 100 match game which ends in 50-50.


This is such an unlikely scenario that the rating system has not been
designed to cope with it. Remember that FIDE Rating periods are only
three months long so it's very difficult to play 100 rated games within a
single period.

Also, your assumed K=20 is a little out -- FIDE uses K=15 for established
players (those who have played more than 30 games) and K=10 for estab-
lished players whose rating has ever been over 2400. Even K=15 will give
the unfair results you describe for this match, of course.

Dave.

--
David Richerby Salted.com (TM): it's like an
www.chiark.greenend.org.uk/~davidr/ E-commerce portal but it's covered
in salt!


  #5   Report Post  
Old August 31st 05, 03:46 AM
Angelo DePalma
 
Posts: n/a
Default


The problem with your scenario is the length of the match.

Had they played 20 separate matches of 5 games each, and submitted a rating
report after each one, both would have a rating of about 1850, which is
halfway between 1700 and 2000. That's ignoring any bonus factors (USCF
awards bonuses for higher-than-expected results).

It strikes me as unfair that this should happen though. Obviously either the
1700 player is much stronger than his rating, or the 2000 player has
suffered brain damage.

I believe this scenario is quite possible, although unlikely due to the
length of the match, between an improving 1700 youngster and a 70-year-old
2000 player. USCF awards rating floors which are 200 points below one's
highest rating class (1700, 1800, 1900, etc.). However, I like Bruno's idea
from his country's federation. There should be some mechanism for
substituting performance rating for actual rating (especially w/ respect to
calculating opponents' new ratings) in situations where one player is
rapidly improving.


"JP Hendriks" wrote in message
...
Suppose player A has a rating of 1700, player B has a 2000 rating.
They play a 100 match game which ends in 50-50.
This is obviously a great result for A and A can be considered about
equally strong as B. But wat happens to the ratings.
According to the FIDE expectation table A was supposed to score 15
points. Assume a K factor of 20 then his new rating is:
Rnew = Rold + K(Realized-Expected)=1700+20(50-15) = 1700 + 700 =
2400
For B the rating result is:
Rnew = Rold + K(Realized-Expected)=2000+20(50-85) = 2000-700 = 1300

This is definitely not a fair result. A now playes at grandmasters
level, and B has become a beginner.

Is there something wrong in the rating calculation; is this a "bug" in
the formula's. Is this problem covered anywhere. I love to get the
option of some rating experts here.

Thank you,
JP.



Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Taylor Kingston caught lying again, says that his Elo rating was 2300+ Sam Sloan rec.games.chess.misc (Chess General) 173 June 30th 05 02:45 PM
Taylor Kingston caught lying again, says that his Elo rating was 2300+ Sam Sloan rec.games.chess.politics (Chess Politics) 197 June 30th 05 02:45 PM
Anti-Cheating Policy HB Super Event rec.games.chess.politics (Chess Politics) 4 March 14th 05 12:43 AM
UPHILL RATING? Squark rec.games.chess.politics (Chess Politics) 26 June 25th 04 02:22 PM
Does unofficial rating of 2200 counts as NM? Denis rec.games.chess.politics (Chess Politics) 8 August 25th 03 03:55 AM


All times are GMT +1. The time now is 10:25 AM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Copyright 2004-2019 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017