FIDE Rating Consultation

[Leonard Barden]

I think I begin to understand what is happening here.

At the Scheveningen tournament where Nguyen Thi Phuong Anh scored 7/9, https://chess-results.com/tnr796350.asp ... =VIE&snr=9 one team was 10 rated players and the other 10 unrated, so the rated players didn't lose any rating points. The rated players did badly, and one WFM didn't win a single game against 8-year-olds. So that created a core of overrated 8-year-olds, and the current tournament starting today is helping to distribute the rating points to others. It's a variation of Myanmar, and presumably legal under current Fide rules. Or is it?

[Roger de Coverly]

At the Scheveningen tournament where Nguyen Thi Phuong Anh scored 7/9, https://chess-results.com/tnr796350.asp ... =VIE&snr=9 one team was 10 rated players and the other 10 unrated,

Stewart's Chess Organiser's book, last copyright 2006 is online at the ECF site.
https://www.englishchess.org.uk/wp-cont ... n-word.pdf

Scheveningen events in which unrated players participate are not rated by FIDE, nor can such events lead to title norms.

But that may have been overturned more recently.

[Chris Rice]

I think I begin to understand what is happening here.

At the Scheveningen tournament where Nguyen Thi Phuong Anh scored 7/9, https://chess-results.com/tnr796350.asp ... =VIE&snr=9 one team was 10 rated players and the other 10 unrated, so the rated players didn't lose any rating points. The rated players did badly, and one WFM didn't win a single game against 8-year-olds. So that created a core of overrated 8-year-olds, and the current tournament starting today is helping to distribute the rating points to others. It's a variation of Myanmar, and presumably legal under current Fide rules. Or is it?

Given their age Leonard it might be the case that some girls would like to draw with each other than have a winner and a loser. I've come across it a few times when teaching girls of primary school age and it can be difficult to explain to them that they are well ahead on material and should be trying to win. However, quite often the response is "but she's my friend!" Of course, it is equally plausible that there are some shenanigans going on as well.

[LawrenceCooper]

I think I begin to understand what is happening here.

At the Scheveningen tournament where Nguyen Thi Phuong Anh scored 7/9, https://chess-results.com/tnr796350.asp ... =VIE&snr=9 one team was 10 rated players and the other 10 unrated, so the rated players didn't lose any rating points. The rated players did badly, and one WFM didn't win a single game against 8-year-olds. So that created a core of overrated 8-year-olds, and the current tournament starting today is helping to distribute the rating points to others. It's a variation of Myanmar, and presumably legal under current Fide rules. Or is it?

I suspect that the tournament will come to the attention of the relevant commission.

[NickFaulks]

Re what defines deflation/inflation: the FIDE Handbook until recently had the following: "It is a major objective to ensure the integrity of the system so that ratings of the same value from year to year represent the same proficiency of play." (FIDE Rating Regulations effective from 1 July 2017 till 31 December 2021 (with amendments effective from 1 February 2021) (https://handbook.fide.com/chapter/B022017), section 10.2.) (Deleted in the latest version; I wonder why?)

I deleted that because it is meaningless and, if it were to have any meaning, would be unachievable.

If Earth went through some meteor shower overnight and we all woke up with similarly improved chess proficiency, ratings would not be affected one jot. Nor should they be. Chess ratings measure how well you score against the player sitting on the other side of the board, nothing else.

[Roger de Coverly]

Chess ratings measure how well you score against the player sitting on the other side of the board, nothing else.

FIDE awards titles based on ratings. Should there not be some consistency of standard in the award of titles over time?

[NickFaulks]

https://ratings.fide.com/profile/12433390/calculations

Vietnam is now showing from nowhere five U9s in the world top 10, while our double world champion Bodhana has dropped from second to eighth. It all sounds a bit Myanmarish, but I've no idea how it is done as presumably the WIM and the WFM are genuine players.

This was noticed by QC and will not be ignored, but there are major differences from Myanmar ( which remains not fully fixed after efforts were scuppered by political intervention ).

Unlike Myanmar, Vietnam has juniors who are very dangerous and regularly chop up much higher rated opponents when they travel abroad. They have been quoted as the most underrated group in the world, so perhaps these results are just as they appear. They do look odd though, and subsequent results will be monitored.

[NickFaulks]

But that may have been overturned more recently.

Yes it has.

A lot has changed since 2006 and it is always worth reading the current regulations, which are readily accessible at https://handbook.fide.com/

[Kevin Thurlow]

"Given their age Leonard it might be the case that some girls would like to draw with each other than have a winner and a loser. I've come across it a few times when teaching girls of primary school age and it can be difficult to explain to them that they are well ahead on material and should be trying to win."

Yes - I've seen that too.

[NickFaulks]

FIDE awards titles based on ratings. Should there not be some consistency of standard in the award of titles over time?

If you could offer a definition of "consistency of standard" that would be a start.

[Roger de Coverly]

If you could offer a definition of "consistency of standard" that would be a start.

There are various possible standards. One if a suitable test could be devised is the level of knowledge required to maintain a particular rating or relative ranking. Over time I think it's increased as chess knowledge has expanded. You can also measure the distance between the world's highest rated player and the standards for various titles. That's widened, but perhaps the rating differences have stretched.

Whether achieved or not, is it not an objective of a rating system to be able to compare the strength of players separated by time, who would never be able to settle head to head who was better?

[NickFaulks]

One if a suitable test could be devised is the level of knowledge required to maintain a particular rating or relative ranking.

Can we leave you to devise such a test?

You can also measure the distance between the world's highest rated player and the standards for various titles.

The top rating is a random outlier and so unsuitable, but I did once make a similar suggestion, that the performance rating required for a GM norm might be linked each year to, say, the rating of the #200 player. I thought this was quite sensible but it found zero support.

Whether achieved or not, is it not an objective of a rating system to be able to compare the strength of players separated by time.

I have never seen the point of objectives which are known from the outset to be unachievable and/or meaningless - I am clearly not cut out to be a politician. Nonetheless, this notion has indeed lurked at the back of ratings discussions within FIDE for decades and has proved a barrier to efforts to improve the system in areas where it actually can work.

[SeanCoffey]

...
I have never seen the point of objectives which are known from the outset to be unachievable and/or meaningless - I am clearly not cut out to be a politician. Nonetheless, this notion has indeed lurked at the back of ratings discussions within FIDE for decades and has proved a barrier to efforts to improve the system in areas where it actually can work.

I had interpreted "from year to year" in the previous language as referring primarily to the short term, for example consecutive years. Rather than comparing different players in different eras, we could compare the same player over short periods of time. For players who have reached a stable plateau, I think most people would expect the rating to remain roughly the same.

That's vague, and (even if a problem seems to be present) it is not obvious what to do about it.

Here is one suggestion. The rating system is, and always has been, a balance between simplicity and accuracy. It might be worth running a more complicated, but probably more accurate, variation of the system, and comparing results.

For example, consider a player who gains 175 points in a single month, going from 2100 to 2275. (This is based on a concrete example, but I'll omit the name.) All this player's opponents are treated as having played a 2100, whereas it seems plausible that they were instead playing a player of rating 2275 or so.

So: variation A: re-run all rating calculations for the year, but with ratings updated on a game-by-game basis instead of the current fixed-ratings-at-beginning-of-tournament model.

Variation B: re-run all rating calculations for the year, but using end-of-tournament ratings for each opponent.

If the median rating of active players changed significantly under one or both variations we would have evidence that the simplification was causing a significant change in ratings. This might provide the impetus to switch to a more complicated system. Conversely, if there is no change, then there is no point changing (and it would provide evidence that there is no net 'deflation', at least attributable to this cause).

[Ian Jamieson]

The rating system is, and always has been, a balance between simplicity and accuracy. It might be worth running a more complicated, but probably more accurate, variation of the system, and comparing results.

I accept I’m in a minority but in my opinion the balance is already too far towards largely spurious accuracy and we should be simplifying the system not making it more complicated. Players and administrators expect ratings to do more things than they are capable of doing, at least for the average player. It would be better if players and administrators were educated on the limitations of ratings and concentrated on their chess not their rating.

[Brian Valentine]

..Access to the results would be good. At present, only Jeff Sonas seems to have access.

I was going to say that one can do quite a bit with just the rating lists. Since I have most of the code I've done a little feasability study. Here's a preliminary conclusion.

I fear Roger's idea of analysing point changes turns out to be a "red herring". In the earliest January standard play list available the average rating is 1802, by 2023 this a dropped to 1628. The downloaded list has very slight changes to the graph impression on page 2 of the Sonas report. This list includes inactive players who don't contribute to the problem Sonas outlines. If one looks at the active players, the earlier average becomes 1781 moving to 1625. In both cases ostensibly serious deflation but almost totally explained by lower rated (not necessarily underrated) new entrants.

While probably what Sonas sets out is related, it is, for the want of a word, expansion. Sonas sets out that problem, but it's not the deflation that can be diagnosed by the movement of points.