Elo FIDE inflation... or not?

[Paolo Casaschi]

Saw somewhere this academic paper that attempts to determine the rate of Elo FIDE inflation over the years.

Simplifying their approach, they took a number of games from FIDE rated players over the years and analysed them with Rybka (assumed to be a better game judge than the players involved) in order to determine whether a 2400 player from the 1979 blunders more or less often than a 2400 player of 2009.

http://www.cse.buffalo.edu/~regan/paper ... eHa11c.pdf

Their surprising conclusion is there is no rating inflation (if anything a slight deflation).

Interesting reading, the description of the statistical analysis is rather complex, but the introduction/conclusion parts are a nice reading.

[Stewart Reuben]

I have had papers presented which demonstrate that there has been very little inflation in the FIDE Ratings this century at the higher levels.
A third gave very convincing arguments that it would be surprising if there had NOT been the growth in numbers of players rated over 2700 that we have seen in the last decade.
I attempted some evaluation of top players in the 1980s. I took established active players in the top 100 and compared their ratings years by year. At that time it looked like about 5 points per year. But, although established (perhaps over 25), they may still have been improving. Also it is a very small sample.
This issue is also discussed elsewhere on this forum and on others.

[Stewart Reuben]

Just as I was going to bed, I realised there may be a serious flaw in the paper of Kevin Regan (of the World Bank).
It considers blunders over the years. But most games in the early years that are on chessbase were played with adjournments. Surely there would have been fewer blunders in say the first three moves in the second session? Thus it is not comparing like with like.
People have asked whether the faster time controls leads to more bad play. This must be so at the very fast rates. Anecdotally, I believe that the cumulative mode causes players to use their time more efficiently. Thus 40/90 + 30 minutes + 30 seconds per move typically lasts 5 hours for 60 moves. 40/2 all in 1 is 6 hours and the play is no better.

[Mick Norris]

I assume the author is IM Ken Regan, who presumably knows what he is talking about
http://ratings.fide.com/card.phtml?event=2000725

[Roger de Coverly]

Just as I was going to bed, I realised there may be a serious flaw in the paper of Kevin Regan (of the World Bank).

I would doubt a serious flaw, not so many games were adjourned. You would exclude games over 40 moves to find out.

The paper is putting forward the hypothesis that players have got better as evaluated by Rybka and the increases seen in the ratings indicate that FIDE has managed this without inflation.

[E Michael White]

The last time I looked at this I came to the conclusion FIDE was deflating at the same time as spreading. At the top level these forces were roughly in balance making the grades at the top reasonable. Spread is more activity related, whereas deflation is more related to time. I expect Fritz/chessbase etc hiked up the strength of players.

Most of the surveys on ratings I have seen rely too much on statistical theory. This is difficult to justify as there is no adequately defined uniform scale to make observations against unlike 100 metre times or earthquake estimation.

Just as I was going to bed, I realised there may be a serious flaw in the paper of Kevin Regan (of the World Bank).

Are you confusing Kevin Regan with Ken Rogoff ?

Under present world banking conditions, It would be a brave person who would suggest the skills of bankers have anything to offer here.