Back to Home Page

FIDE mess up January Ratings (big time)

FIDE rating page

Jan 6: FIDE will have to withdraw the January 2003 rating list after a disastrous error in the processing of games from the BLED OLYMPIAD in October-November.

Jan 9: FIDE have acted promptly to withdraw the initial rating list and eliminate all the erroneous games against unrateds. The Collins and Williams errors highlighted below both seem to have been fixed. Amazing what a link on TWIC will do for getting grading errors amended speedily!

Of course it wouldn't be FIDE if it was 100% accurate, Sam Collins (SCO) is still not in the new download, but does appear in the itemised player lookups. All the players in the Edinburgh all-play-all should all be adjusted downwards, since the rated Sam Collins (IRL) was incorrectly in version one and thereby boosted the average opposition of all players. So all these ratings are still wrong. There is still no indication at the FIDE site that they have changed the initial list.

It would be interesting to see a complete breakdown of where all FIDE finances (ie the cash from federations) go on rating. How many paid FIDE employees work on the list for example. Some of that federation cash should go on developing software to give to federations for standardised rating reports which FIDE can import directly into their system (see below).

Heather Lang, pictured right, was one of the first players to notice that she had far more games processed than expected. FIDE had blundered by including all games against unrated opposition. All unrated players were given a nominal 2000 rating. Many players at Olympiads, especially in the Women's competition, are of much lower strength than 2000. The error also occurred in the Men's event. There were 135 players in the Open (Men's) Olympiad and 97 unrated in the Women's - presumably all counting incorrectly as nominal 2000. Anyone who played one of these unrateds will have an incorrect January rating. Approximately 2100 individual game results (two results per game) were affected.

Email from Heather on Jan 3:

"Looks to me like they've made mistake grading the women's Olympiad (this is from looking at the calculations for me and for some of my opponents).

They've taken the ratings of my rated opponents, and counted my unrated opponents as 2000. They've then given me 7/10 against them all and calculated my increase on that basis. As I got 4.5/5 against the unrateds it boosts my rating far more than it should (the correct figures are given in the final round bulletin).

For example, Debbie Quinn of Ireland, one of my unrated opponents, scored 1/5 against an average rating of 2143. Somehow she's arrived on the rating list at 2036. FIDE's website has now gone down, but I suspect that she's been given credit for results against unrateds counting as 2000 (total score 4.5/10 vs 2072 if this is the case) - I don't think she's played in anything other than the Olympiad in this rating period.

I doubt there'll be a huge outcry as it won't affect the top players, just
those who played unrateds, and as people are getting credit for it,
they're likely to keep quiet. :-)"

I passed Heather's email to FIDE official Stewart Reuben. "Do unrated players count at the Olympiad?"

SR: "They have made a ridiculous error.
Of course unrateds should not count as 2000. They were counted as 2000 for the purpose of seeding. That is a decision Geurt and I made and it was logical to do so rather than 1800. Unrateds count in no way different from in a normal tournament.
That is a serious mess and FIDE should be asked to correct it immediately. I will write."

And Casto Abundo, rating officer in Elista, thinks that Guert Gijssen, chief arbiter in Bled, is to blame:

Dear Stewart

You are right. Due to a mistake in the report of the Chief Arbiter, the Olympiad ratings are wrong.

C. Abundo

What a mess. FIDE are incapable of producing ratings in accurate fashion. There are constant errors in player identity. I cannot recall a production of the rating list when they did not have to be contacted about some error or another.

The Edinburgh FIDE rated tournament in October featured SAM COLLINS of Glasgow who did enough to get a new rating in the January 2003 list. He is not in the list since FIDE decided to replace our Sam Collins with the star Irish junior of the same name, this is despite Chess Scotland listing Collins as - unrated - SCO for Scotland - and we even gave his date of birth in 1951. How on earth can FIDE think this is the same person as someone who has just played in the world junior championships in India, restricted to players under 21.

If FIDE had any doubt about this player why did they not contact Chess Scotland or the Irish federation to clarify?

John Saunders at the BCM reports a similar error has occurred with Leighton Williams of Wales having all his games given to some Colombian.

Instead of wasting time on thinking up player humiliations like peeing into bottles for drug tests and amending rules like "take king loses at blitz" (they actually did this)... you might think FIDE could put a bit more effort into something as fundamental as getting ratings produced accurately.

One simple step to improve FIDE rating would be to supply federations with standard software to submit results. It is incredible that this basic point still needs to be addressed. FIDE complain about most data still being submitted on paper forms - well, no wonder. Why not create a program and you solve the problem immediately. A federation about to submit data would get a copy of the latest database directly from FIDE and the program to submit results would select the names directly from the database. Therefore CS would not have selected the rated, Irish, Sam Collins from the database and it would be clear to FIDE this was a new player.

IA Alex McFarlane had to fiddle about with an Excel spreadsheet to try and get an accurate report of the Edinburgh FIDE event. An A-P-A report is a complete nightmare to produce, try following the instructions in the FIDE handbook without wanting to throttle someone. Of course you shouldn't have to suffer this nonsense, a program should take care of the arithmetic. Why don't FIDE take the hassle away from federations by producing software that deals with this - at the same time they standardise reporting and make their own life much easier.

FIDE thought extending the rating list downwards was going to be a big money spinner since federations could be charged more. As they go down the scale to the ultimate aim of 1001 the base of the pyramid gets wider and the task is going to get more and more difficult.

Being a rating officer is a thankless task since you get little praise when things go right and heaps of blame when errors occur. Chess Scotland since 1997 have produced itemised result reports on the website of every game processed for grading. I recall Sarah Hurst, then of the BCF, complaining about how boring were all these result lists on the Scottish site. However it must be quite a novelty to have a federation which tries to produce a rating that broadly coincides with games actually played.

There are errors but the players themselves can see exactly what has been processed and report the error before grading lists are published. The latest facility by programmer Alex Bissett of Aberdeen allows players to look up exactly what is on the grading computer - online lookup. FIDE will need a similar service if they are ever to prevent errors engulfing every list.



Home -   Congress Ads  -  Leagues  -  Grading  -  Chess Scotland Info  -  Schools  - Downloads  -  News  -  Links
Chess Scotland 2002