FIDE: We support the increase of the K-factor

4/29/2009 – Yesterday we published a letter by GM Bartlomiej Macieja asking the World Chess Federation not to delay the decision to increase the K-factor in their ratings calculation. Today we received a reply to Maceija's passionate appeal from FIDE, outlining the reasons for the actions. In addition interesting letters from our readers, including one from statistician Jeff Sonas. Opinions and explanations.

Singapore, 28th April 2009

GM Bartlomiej Macieja

Dear GM Macieja,

We refer to your email to FIDE and the open letter which was published on ChessBase on 28th April 2009 concerning your comments on the K-factor.

Firstly, in order not to create any misconception, we would like to assure you that we are in support of increasing the K-factor. However, in view of such a radical change, we would like that all players especially the top players, understand the consequences of the change. We believe that a full proposal from the Qualification Commission should also include changes in the title regulations if necessary, but unfortunately there was no mention as to whether this was necessary. Moreover, the proposal on K-factor was not even highlighted during the verbal report of the QC at the General Assembly of Dresden 2008.

In Dresden 2005, a resolution was made that when the General Assembly/Executive Board could not discuss reports in detail such reports are referred to the Presidential Board to further decide. Attached (below) is the said resolution “Dresden 2005 Resolution Commission Reports”. Hence our decision to delay the implementation is not “to challenge” the reports made at the General Assembly.

We foresaw a problem when FIDE was informed that at the Amber tournament in Nice, France, that all the top players who were participating there had agreed to sign a protest letter (a draft was already prepared) to the FIDE President concerning the change of K-factor. Surprisingly, among these top players were several young players who supported the protest. This proved that the top players had not been duly informed nor consulted on such important changes.

We understand that publishing a list by calculating past years is an indication to the players. However, we felt it was even better to publish a list which includes events of 2008 to 2009. At the same time, from July 2009, to publish the parallel list using the new K-factor so that the players would be better able to understand and feel how they would be progressing in present time than just in the past. Don’t forget Mr Macieja, FIDE and the Continents use the Rating List to qualify players for events which are included in the World Championship Cycle. With the change in K-factor, some players would be in and some would be out. We want to give the players time to realise how the change could affect their career. This decision offers opportunity to all players, federations, organisers and rating experts to give relevant input and comments.

Even if FIDE has the power to force such changes, we believe it was not the best way forward if we are unable to convince the greater majority particularly so since this serious change would affect the lives and career of especially the top professional players.

The comparison of our top chess tournaments with tennis is not relevant. Most of the top-level chess tournaments are round-robin events where a small group of players are invited. Henceforth, any fluctuation in the ratings of the invited players could adversely affect the category of these top-level chess events. In tennis, most players are invited and any fluctuation in their rankings does not affect the event.

In conclusion, we reiterate that both FIDE and the undersigned are in favour of the change. FIDE as well as the ACP should communicate these changes and convince the national federations and players.

We take this opportunity to commend you for your efforts and that what you have contributed towards rules and regulations in FIDE have been positively considered.

Best Regards,

Signed : Georgios Makropoulos, FIDE Deputy President
Ignatius Leong, FIDE General Secretary


Annex 48
(1 page)

Proposal for the Executive Board following approval by the Presidential Board

1. Proposals tabled at the Annual Congress without due notice shall be subject to appraisal at the first Presidential Board meeting following the Congress. The Presidential Board shall not have the power to rescind the decision, only to delay implementation until the proposal is further studied at the next Congress.

Reason: Each year at Congress we face situations where papers from Committees and Commissions are presented for approval without adequate time to digest the full statutory or financial implications of any proposals in the documents. There ought to be a mechanism to allow further reflection. Committee Chairmen should be encouraged to table an agenda for their own meeting at Congress in time for it to be circulated with the full Congress agenda issued by the Secretariat. They could highlight any proposals that they intend to have discussed that might have financial implications or require changing regulations. This would give adequate notice of potentially contentious issues and allow the Treasurer to have some input where there were budgetary consequences.

Extract of minutes

Mr. Jarrett said that the second proposal relates to the fact that in all meetings of the FIDE Congresses, we receive many papers and it is not possible to fully digest and understand all the consequences. The request is that all decisions passed can be referred to the next Presidential Board to be properly examined. This does not give the Presidential Board a right to waive the decision taken, but it will have a chance to present a different angle for the next General Assembly.

Mr. Makropoulos said that the issue is about the late submissions, coming from Committees and Commissions during the Congress.

Executive Board approved the proposals of Mr. Jarrett.


Letter from chess statistician Jeff Sonas

I was interested to see the recent announcements regarding official changes to the FIDE rating system. My analysis in 2002 recommended four significant changes to the FIDE rating calculations:

  1. Using a more dynamic K-factor (24 instead of 10) would result in ratings that more accurately predict players' future results, and thus I would call those ratings more "accurate".

  2. We should discard the existing arbitrary scoring probability table (see FIDE Handbook, 10.1.b) that is used to recalculate players' ratings, in favor of a simpler linear model that does not discriminate against higher-rated players.

  3. Including games played at faster time controls, although at a lesser weight than regular games, would provide additional useful information for a more accurate rating calculation.

  4. Calculating the ratings more frequently than just quarterly is easily done and would presumably improve the accuracy of the rating calculation.

It appears that FIDE has decided to move in the direction of my recommendations #1 and #4, since they are going to use a K-factor of 20 (or 30 for players who have never reached a rating of 2400+) instead of 10, and they are going to release official rating lists six times a year rather than four times a year. In both cases, my 2002 analysis recommended a more significant change, but nevertheless I do like the direction this is going.

From the notes published by GM Bartlomiej Macieja, I see that there is some controversy regarding:

(a) the impact these changes would have on rating inflation
(b) what difference it makes mathematically whether ratings are calculated yearly, quarterly, monthly, etc.
(c) the decision-making process for how the rating calculation process will change.

The question of rating inflation is a very difficult one and really the only way to tackle it is to look at actual data and see the result of various approaches. I believe that with a more accurate rating formula, such as what I have suggested in my analysis, you could see significant rating inflation, even more than we currently have. The natural inflation in the Elo approach is currently being offset by biases in the Elo expectancy table against the higher-rated players. So the current formula does limit inflation, but at the expense of players who play against lower-rated opponents. There are fairer approaches you can take to halt inflation, such as periodic small reductions to everyone's rating by the same amount. That would take some getting used to, but would eliminate inflation fairly if that is indeed a primary concern.

I have been limited in my past analysis by not having the official FIDE historical data that was used in the past to recalculate ratings. All I can do is to analyze against a test dataset of games taken from a source like TWIC or ChessBase's Mega Database. This provides very useful information but doesn't necessarily give you the whole picture about the impact on all players, including lower-rated players. If FIDE would like to provide me with the official data used in the past historical calculations, going back at least 5-10 years, I would be happy to perform another analysis, in hopes of helping to understand better these questions about rating inflation and frequency of rating calculations. I'm afraid I can't help much with the problems about the decision-making process, nor do I have any interest in tackling that! But I'm sure I could help answer questions about the mathematics of the rating formula, if anyone is interested...

Jeff Sonas

Jeff Sonas is a statistical chess analyst who invented the Chessmetrics system for rating chess players, which is intended as an improvement on the Elo rating system. He is the founder and proprietor of the Chessmetrics.com website, which gives Sonas' calculations of the ratings of current players and historical ratings going back as far as January 1843. Sonas writes that, "Since the summer of 1999, I have spent countless hours analyzing chess statistics, inventing formulas and other analysis techniques, and calculating historical ratings." He has written dozens of articles since 1999 for ChessBase.com and other chess websites.


And while we are at it, here are some letters we received in reaction to Bartlomiej Macieja's article:

Brian Carson, Toronto, Canada
Comments on the ratings calculations by GM Bartlomiej Macieja is nothing more than a voice of reason in a frothy sea of emotional irrationality. I wholeheartedly agree, increase the "K" and start enjoying reality, life is too short to pretend you are something you are not. Many epitaphs may read "he was briefly a 2700 player" but reality always wins. It is called checkmate.

Gerd Lorscheid, Boeblingen, Germany
The slowdown of the rating change shown by Macieja is correct. But it happens only if the player wins rating points in all his tournaments. On the other side there is a risk with K=30. If a player with rating 2350 needs 2400 to get the IM title, he can do the following: He lets drop his rating to 2200, play in one period 40 games with a performance of 2350. His next rating will be 2425!

David Fournie, New York, USA
Macieja's comparison with Tennis for invitation to tournaments makes no sense because the ATP points won by participating in a tournament does not depend on the ranking of the other players (to be more precise, the number of points you earn in function of the tournament stage in which you are knocked out is fixed in advance at the time the tournament is announced, independently on who plays in the tournament). For a chess tournament, the rating points you gain or lose for a given score depends on the average rating of the opponents hence a player would always lose points if he plays in a tournament with a much lower rated opposition. This is why ratings fluctuation should not be to wide (and hence K-factor not to big), so that when one accepts an invitation almost a year in advance he is pretty confident about the average ELO of the opposition.


Feedback and mail to our news service Please use this account if you want to contribute to or comment on our news page service



Discuss

Rules for reader comments

 
 

Not registered yet? Register