PDA

View Full Version : Game by Game ratings, fide



Garvinator
04-10-2005, 12:00 PM
This item has popped up on fide's website:

http://www.fide.com/news.asp?id=820

Each game counts separately.

FIDE approved the system of game by game rating, a definite improvement over average opponent rating generating more accurate statistical results. A standard format for reporting was approved by the Dresden FIDE Congress. Designed by FIDE Swiss System Committee Chairman Christian Krause, the format is being followed by FIDE-endorsed programs such as the free software PROTOS. All reports must be in this standard format by July 2006. Until then, some reports shall show only averages.

An advantage of the new format is the more detailed presentation on the FIDE web site for all players to view their game by game calculations with a list of their opponents. For example, see Topalov`s results in Dresden Sparkassen.

Moreover, the system of openness and more detailed presentation contributes to the credibility of the FIDE ratings. Transparency shall help prevent fraud not only for calculation of ratings but also for title applications. The increasing number of FIDE rated players, over 64,000 now, requires more checks and controls.

Note that we publish birthyear instead of birthdates for security of personal information. We continue to update our database. National federations may request and receive a copy of the FIDE Rating List with full birth date information as this helps in identifying players and discrepancies between federation and FIDE data.

We hope to continue improving our service and appreciate your comments and suggestions. Please send corrections, if any, to ratings@fide.com.


Now I assume that Bill might have something to add to this.

bergil
05-10-2005, 06:51 PM
This item has popped up on fide's website:

http://www.fide.com/news.asp?id=820

Each game counts separately.

FIDE approved the system of game by game rating, a definite improvement over average opponent rating generating more accurate statistical results. A standard format for reporting was approved by the Dresden FIDE Congress. Designed by FIDE Swiss System Committee Chairman Christian Krause, the format is being followed by FIDE-endorsed programs such as the free software PROTOS. All reports must be in this standard format by July 2006. Until then, some reports shall show only averages.

An advantage of the new format is the more detailed presentation on the FIDE web site for all players to view their game by game calculations with a list of their opponents. For example, see Topalov`s results in Dresden Sparkassen.

Moreover, the system of openness and more detailed presentation contributes to the credibility of the FIDE ratings. Transparency shall help prevent fraud not only for calculation of ratings but also for title applications. The increasing number of FIDE rated players, over 64,000 now, requires more checks and controls.

Note that we publish birthyear instead of birthdates for security of personal information. We continue to update our database. National federations may request and receive a copy of the FIDE Rating List with full birth date information as this helps in identifying players and discrepancies between federation and FIDE data.

We hope to continue improving our service and appreciate your comments and suggestions. Please send corrections, if any, to ratings@fide.com.


Now I assume that Bill might have something to add to this.
Not yet he hasn't and if adds work for him he would do well to ignore it.

So whats the standings in the GP this year, who are the leading players? Where is a list of all the tournament results? Are the numbers of players up or down? why is it not in the ACF newsletter as a monthly report with standings of all divisions?

bergil
08-10-2005, 01:01 AM
So whats the standings in the GP this year, who are the leading players? Where is a list of all the tournament results? Are the numbers of players up or down? why is it not in the ACF newsletter as a monthly report with standings of all divisions?
Whoa look out for the tumbleweed!! :P

jase
08-10-2005, 01:14 AM
Bergil this is a somewhat devious tactic, to subvert gg's thread into the much-maligned GP :buttkick:

For people like myself, Shaun Press, and Brian Jones, who have previously invested a lot of time and energy into the GP, it's unfortunate to see it lying prostrate in a gutter.

As for the FIDE ratings - an interesting development. I too would be interested to know Bill's thoughts on the matter.

bergil
08-10-2005, 01:23 AM
Bergil this is a somewhat devious tactic, to subvert gg's thread into the much-maligned GP :buttkick: Nah not really, just love all the comments and suggestions from GG for others



For people like myself, Shaun Press, and Brian Jones, who have previously invested a lot of time and energy into the GP, it's unfortunate to see it lying prostrate in a gutter.

You all did good work, yes it is but maybe it will turn around next year :pray:

Rincewind
08-10-2005, 08:43 AM
For people like myself, Shaun Press, and Brian Jones, who have previously invested a lot of time and energy into the GP, it's unfortunate to see it lying prostrate in a gutter.

It is disappointing but sometimes one has to lie prostrate in a gutter to help one appreciate the good times. Here's to happier times for the GP!

bergil
10-10-2005, 03:08 AM
It is disappointing but sometimes one has to lie prostrate in a gutter to help one appreciate the good times. Here's to happier times for the GP!
So it would be a good idea for the state bodies to get behind the revamped 2006 GP series, don't you think Rincewind, Jase, anyone? :hmm:

pax
11-10-2005, 01:33 PM
Each game counts separately.

Interesting development, and about time. I've been talking about this idea for about 15 years (it is, after all the way Elo envisaged it - batching was only done to make it computable). This removes the artificial discontinuity on rating changes over whether they appear at the start or end of a rating period.



Now I assume that Bill might have something to add to this.

Bill will tell you that Glicko is designed to be batched, and he is right.

IMO it is one of the weaknesses of Glicko that it doesn't like to be "game at a time" batched.

Game at a time rating would also prevent bill from doing his 'intermediate rating' thing.

four four two
11-10-2005, 02:10 PM
And maybe one day FIDE will be competent enough to have the ratings updated fortnightly,but that would mean actually employing someone full time. :doh: :pray: :idea: :whistle:

Garvinator
11-10-2005, 02:18 PM
Game at a time rating would also prevent bill from doing his 'intermediate rating' thing.
If game at a time rating was done then there would be no need for an intermediate rating as each player would receive a 'new' rating after each game. Therefore the 'under-rated' juniors would be receiving a new rating after each game during the three month period when calculated. Of course this also means they can go down faster.

Bill Gletsos
13-10-2005, 04:09 PM
If game at a time rating was done then there would be no need for an intermediate rating as each player would receive a 'new' rating after each game. Therefore the 'under-rated' juniors would be receiving a new rating after each game during the three month period when calculated. Of course this also means they can go down faster.It is clear that you have no clue what you are talking about.

In fact you and others dont seem to even understand what FIDE means by game by game rating.

FWIW examination of the FIDE website would seem to show that for the October 2005 list some events were calculated on a game by game basis and other events were done on the old average rating of opponents method.

four four two
13-10-2005, 04:37 PM
So why are they rating some events differently Bill? Was the change made during the middle of the rating list period? :hmm:

Bill Gletsos
13-10-2005, 04:44 PM
So why are they rating some events differently Bill? Was the change made during the middle of the rating list period? :hmm:I can find no explanation as to why they are done differently.

As far as I can tell the change was approved at the recent FIDE Congress in Dresden in August.

However the Dortmunder Sparkassen tournament run in mid July was calculated based on the game by game basis rather than the average of opponents rating basis.

1min_grandmaster
13-10-2005, 06:32 PM
I don't quite understand the difference between 'game-by-game rating' and 'average opponent rating'. Can someone provide a concrete example with actual numbers showing the difference between the old and the new system?

Bill Gletsos
13-10-2005, 08:51 PM
I don't quite understand the difference between 'game-by-game rating' and 'average opponent rating'. Can someone provide a concrete example with actual numbers showing the difference between the old and the new system?Ok here is how it works.

An established player is rated 2650 at the start of the rating period.
He plays 7 games with the following results:


Rating Result
2300 0.0
2450 1.0
2450 1.0
2550 1.0
2400 1.0
2465 1.0
2486 0.0
Now for the old method of average opponents rating the calculation is as follows:

Average rating is 2443.
Therefore rating difference is 2650-2443 = 207
Expected winning percentage for a difference of 207 is 0.77.
The expected score is 0.77 * 7 = 5.39
Actual score is 5.
Difference in actual - expected = -0.39
Rating change is -0.39 * K
Since player is rated over 2400 K = 10

Therefore rating change = -0.39*10 = -3.9


Now for the game by game method it works as follows:


Rating Diff Expected
2300 350 0.89
2450 200 0.76
2450 200 0.76
2550 100 0.64
2400 250 0.81
2465 185 0.74
2486 164 0.72
Now sum of expected = 5.32
Actual score is 5.
Difference in actual - expected = -0.32
Rating change is -0.32 * K
Since player is rated over 2400 K = 10

Therefore rating change = -0.32*10 = -3.2

Also it should be noted that the 2650 players rating remains 2650 for all his games during the rating period even in other events during the rating period. The only exceptioon is if the event rated in this period started prior to his 2650 rating being published.
An actual example of this is the current San Luis event. The calculations are all done with the players rating based on their July 2005 FIDE ratings not their October 2005 FIDE ratings even though the event will be rated in the Jan 2006 FIDE list.

Players ratings are not adjusted after every game nor are they adjusted after every tournament.

The FIDE game by game calculations are done tournament by tournament but the result would be the same even if it was done on game by game for the whole period. It is done tournament by tournament to allow the players to see how their rating is affected by their results on a tournament basis. No doubt because the players are used to seeing this already.

Note that with the average opponents rating method FIDE also does the calculation on a tournament by tournament basis but if they did it on the average opponents rating for the whole rating period the calculated rating change could well be and would most likely be different.

1min_grandmaster
14-10-2005, 11:11 AM
Thank you very much for your clear example Bill. For some reason, I just assumed that the FIDE ratings had always been calculated in the 'obviously correct', game-by-game method.