3

The background is covered in this question and the accepted answer. The changes were delayed until the March list. Comparing charts for the number of different players at each rating level before and after the anti-deflation measures, what does this look like?

1 Answer 1

6

I took the FIDE rating data from the FIDE rating download page, loaded it into a database and ran some queries to get scatter plots for February 2024 and March 2024.

This is what I got for February:

enter image description here

and March:

enter image description here

Interesting to note that the historic rating peaks at 2005 and 2205 are still there from the days when FIDE only rated women above 2000 and men above 2200.

However, what raises a whole load more questions is the smearing between 400 and 900 on the Y (player count) axis. This suggests something strange about the way the anti-deflation calculations were done.

Illuminating comment from DM:

Due to the formula used, some new ratings got two old ratings, and some got only one. 1186/1187 both got converted to 1512, and 1188/1189 both got converted to 1513, but only 1190 got converted to 1514, for example

Since the April data is also available now I tried doing the same with that and this is what I got:

enter image description here

The smearing seems to have got a bit narrower but this has another anomaly. There are a very small number of players between 1266 and 1400. They shouldn't be there because there is now a rating floor at 1400. I chased down one or two of these players and here is part of the download data for a player listed as rating 1280 -

Download Data
fideid 423440 name Shoulders, Andrew M country ENG sex M rating 1280 games 8

Mr Shoulders FIDE rating progress chart is here and he is now correctly shown as unrated having been 1418 in March. Which suggests the FIDE system may still have some teething problems. If you look at his individual calculations for April 2024 you can see what happened.

Note that the main reason for the download data is not to satisfy data geeks but is used by pairing programs like Swiss Manager. When you press the "update player ratings" button the program goes away and downloads this data, parses it and updates the program's rating data.

4
  • To interpret your graph, the smearing means that if I zoom in on the x-axis to for example ratings from 1600 to 1610, then there are between 400 and 800 players at these rating levels. This is odd and probably an artifact of the adjustment because there should be a comparable number of players at almost equal ratings. Maybe with the adjustment even ratings are a lot more likely than odd ratings or something like that.
    – quarague
    Apr 22 at 9:13
  • 2
    @quarague It was every third rating instead of even/odd, but that's more or less correct. Due to the formula used, some new ratings got two old ratings, and some got only one. 1186/1187 both got converted to 1512, and 1188/1189 both got converted to 1513, but only 1190 got converted to 1514, for example.
    – D M
    Apr 23 at 4:21
  • 1
    I think the graphics would be more readable if you used scatterplots rather than line plots (which cause all the almost-vertical lines going up and down between y-400 and y-800 that are blurring the data rendition)
    – Evargalo
    Apr 24 at 9:14
  • im also interested in certain countries also ignoring inactive players
    – cmgchess
    May 12 at 5:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.