09 September 2014

Ratings Deflation

There are several Malaysian chess bloggers or anonymous commentators who try to promote the fiction of 'ratings inflation' at every opportunity. For example, they'd say a 2300 rating 20 years ago is equal to 2400 nowadays. As we say in Malay, that is 'cerita karut'. So I did some research to see how widespread this misconception is.

To my pleasant surprise, the hot topic in chess forums and articles worldwide on this matter discuss ratings deflation that is especially pronounced at the 2000-2200 range. Even FIDE recognises this ratings deflation issue from a long time ago.  Links to forums discussing ratings deflation: wikipedia, rybka forum, chess.com.

I suspect that it is only a few Malaysian chess oldsters who is out to promote themselves by saying that our younger players are weaker than they were a long time ago. Actually, a 2400 rating only 15 years ago is probably about 2350 nowadays. and a 2200 rating say 20 years ago is unlikely to be more than a 2000 rating nowadays.

Inflation only at the top end 
The idea of ratings inflation was first presumed at the top end of the range, when the ratings of the world's best players appeared to be increasing with time. At first glance, it should not be because FIDE rating is supposedly a 'zero sum' system. If a player gains rating points, another player must lose ratings points. So over time there should not be any inflation overall.

However the observation is that there is clearly inflation at the top end  So if the top players' ratings are going up, then someone else must be losing ratings points. The logical conclusion is that there must be ratings deflation elsewhere within the ratings range.

Deflation at the lower end of the range
So, even from a long time ago, those who studied these matters had come to the conclusion that there is ratings inflation at the top end which is balanced out by ratings deflation at the lower end. The question then moves to what I call the 'inflection point', where deflation stops and inflation starts. When I was researching this matter a few years ago, my estimate was an inflection point around 2500.

However, some two years ago, FIDE commissioned the statistician, Jeff Sonas, to do a study using the ratings data provided by FIDE. Mind you, it was 100% data and not just a sample which make it a comprehensive study. Surprise, surprise, the inflection point was found to be just below 2700. This means something like say, comparing ratings over a 5-year period, it could be:

5 years ago       Now       Difference
  2780               2788            8
  2750               2756            6
  2720               2724            4
  2700               2702            2
  2695               2695            0
  2660               2659           -1
  2600               2597           -3
  2500               2495           -5
  2400               2393           -7
  2300               2290          -10
  2200               2185          -15
  2100               2075          -25

Thus, in this example, 2780 five years ago has been inflated to 2788 in today's ratings, On the other hand, ratings deflation has caused a player with 2100 strength five years ago to drop to 2075 now even though there may not be any drop in playing strength. This situation is applicable to a mature player who plays regularly especially against opponents within say a 200 point ratings difference. 30 games a year should be the norm for 'regular play'.

For a player who plays less than 30 rated games a year, the likelihood is that his rating does not reflect current strength. So for a mature player who hardly plays, a 2400 rating 10 years ago could be worth less than 2300 nowadays. And a 2200 rating 20 years ago is probably less than a 2000 rating nowadays

So if an 2200-rated (20 years ago) player who has been inactive (but still play at the strength as 20 years) now comes out to play say 30 games within a one-year period against a mix of opponents in the 2000-2400 range, it is likely that his rating will quickly drop to below 2000 to reflect today's higher standards.

This article is written at the request of Fong Yit Ho who wishes to ensure his contemporaries are not misled by the propaganda of some 'old' Malaysian players who are always saying that they were stronger than the younger Malaysian players at the same age . They do this promoting the fiction of 'ratings inflation' thoughout the whole ratings range which is clearly a falsehood.

3 comments:

  1. I don't know who you refer to saying this rubbish but I can tell you for sure that if they are from my generation none were as strong as the young players at 18 to 21 years of because then we started playing late (I for example learnt chess at 13-14 years of age) and had very few tournaments so irregardless of talent shown during MSSM time, most of us got stronger in our late twenties after getting more experience.

    ReplyDelete
  2. Based on Jeff Sonas' table above, a 2200-rated player from 5 years ago would have deflated to 2185 today, i.e. a 15 point drop. On what basis do you conclude that a 2200 player from 5 years ago should be equivalent to a 2000-rated player today (i.e. a far bigger 200 point drop)?

    Granted I accept your argument that someone who plays regularly should, in practice, be better/stronger than an equivalently-rated player who plays less regularly. However, this is actually an argument on relative strength of actives-vs-inactives, rather than rating inflation/deflation. The rating inflation/deflation phenomenon, which as you pointed out from Jeff Sonas' report, does not appear to have a significantly material impact either way.

    ReplyDelete
  3. To
    Anonymous 2.00 am
    I don't know what you are talking about. Few points:
    1) The table in the post is my creation just to illustrate the inflection point concept.

    2) My personal view is stated in my article, that is there could be a 200-point difference is if we compare now with 20 years ago. You have to prove it if you think I am wrong.

    3) My article focuses on inflation/deflation. Strength of active/inactive is a totally different topic, which I do not wish to discuss.

    You appear to try to put a lot of words into my mouth, as in
    1) "On what basis do you conclude that a 2200 player from 5 years ago should be equivalent to a 2000-rated player today (i.e. a far bigger 200 point drop)? "

    There was nothing of this sort in my post.

    2) "Granted I accept your argument that someone who plays regularly should, in practice, be better/stronger than an equivalently-rated player who plays less regularly."

    I did not make this argument at all.

    I do not wish to get into a debate with you since either you have poor English comprehension, or you are abusing this comment feature to deliberately distort my statements and mislead readers.

    Anyway I am locking up this blog to go into retreat for the next 7 weeks or so. So comments in this blog are disabled as of now.

    Thank you to all readers for visiting and contributing your views is the last few weeks.

    ReplyDelete