Rating Inflation: 100 Elo points starting in 1985

First, read this article: http://en.chessbase.com/post/elo-oddities-the-tortoise-and-the-hare

Next, if you believe in ratings inflation as hypothesized by Sonas in the article, and that today's players are not necessarily more skilled than yesteryear's (in particular, note the jump starting in 1985--did Kasparov really cause a huge jump in theoretical knowledge of chess?), and note the eventual 100 point ratings inflation in Elo.

Finally, go to 2700chess.com and look at the all time Highest Ever Live ratings. 

Now, from the Chessbase article graph make the adjustments: if before 1985 no inflation, if after 1985 up until 1995 the "inflation" is roughly 50 points, if after 1995 the inflation is about 100 points.

So Fischer retains his title as highest rated player ever.

Uninflated ratings (or deflated as the case may be) as per the above:

Fischer: 2789.7 (still #1!)

Carlsen: 2789.2

Kasparov: 2756.7

Karpov: 2740.9

Aronian: 2735.5

Anand: 2720.7

...and so on.  Again, you have to either believe in the Sonas hypotheis for this to work.  If you truly think that today's players, after 1985, are indeed radically better than a generation ago this will not work.  Along this line, some people including GM Nunn have speculated that due to the poor state of knowledge in the old days, indeed 100 years ago the players were in fact much inferior to today's top players.  And you see the same thing in athletics (today's athletes are bigger, faster, stronger than yesteryear's).  But the jump starting around 1985 that eventually added about 100 points to Elo looks suspicious. 

hmmm

I should point out that IM Ken Regan, a computer scientist, has a good presentation that shows there is no ratings inflation after all, you can find it on the net.  But to me the jump around 1985 looks very suspicious.

Very interesting. Good topic

Not really: around 1980 computers started to beat GMs. So probably GMs started to use them for analysis, and thats why they became stronger.

Addon: another explanation is the rise of chessbase (founded 1986) and other chess databases. Also a wider spread of Chess Informator (published since 1966) may be the reason. Finally the end of UDSSR in 1991 may have contributed to a wider spread of chess literature.

Addon 2: And of course the rise of the internet may have contributed.

Numbers often make fools out of those who try to add meaning to them.

The ratings are what they are. They can only be 'inflated' if it is known what they 'should' be. And, since there is no such thing as a set rating value, then there can be no inflation or deflation in rating points.

I also would like to add, that around the eighties chess players started using computers for analysing, and thus came a 'new wave' of chess, if you'd like. So the jump is not at all unbelievable, it is the result of a new approach.
 

I'm sure there was a similar 'inflation' when books like My System were published and everyones game improved.

A counterpoint is a paper by computer scientist D. Ferreira, who cites Regan, that finds 100 Elo point rating inflation (roughly) between 1971 players and today's, though even so Fischer would only be ranked #10 by today's standards and Korchnoi, Petrosian, Hubner would be #12.  See more at this thread: http://www.chess.com/blog/Nimzoditch/test-blog-post-were-old-masters-really-worse-than-present-masters

Even in light of facts and clear evidence people hero worship to try justifying why yesteryear's greats are better than today's!  Lasker, Alekhine, Botvinnik, and Smyslov would be humble enough to admit that today's top players far outclass them.  Even Fischer's penchant towards realism would eventually admit this fact.  Have you seen Carlsen's technique in major piece endings?  With two rooks and a queen vs. similar or rook vs. rook against Anand in equal positions he still manages to find resources and tiny mistakes that would even escape Smyslov's attention! 

FIDE gave all women except for the Polgar sisters a 100 point jump.  Rating floors mean easy points for opponents of people who are playing below their artificially rating.  Both of these would inflate the rating pool, especially at the top.

Could you please elaborate? I had the impression computers reached ~2600 elo in the mid 90's. Was there any software or dedicated chess hardware available in the early 80s strong enough to challenge Karpov or Kasparov?

A general increase in players strength can never explain rating inflation. 

If you're talking about across-the-board - board ratings, including the entire player pool, the only thing that can cause a general increase in average rating is the rating you assign to newcomers in the pool. 

A different thing is if you take only a subset of players, say only the top 100. In this case, this subgroup's rating could be inflated by their improvement in chess skills, but only relative to the rest of the players. 

Barring other reasons for inflation, the fact that Carlsen's rating is higher than Fischer's oye means that Carlsen's skill is more distant to an average player's skill than Fischer's skill was compared to an average player in the 1960s, but it says nothing about Carlsen's skill as compared to Fischer's. 

A general increase in player strength can explain a general increase in rating. 

If whoever played best in 1985 was 2700+ (2-3 players) and if more knowledge, ability and precision exist in the human pool today, you will have more 2700s today, and also a few 2800s.

It's actually not even a general increase in rating - because you still have the 2000s, 1800s, 1300s and lower. People who know less and play worse will still get those low rating. 

The expansion at the top may well be the result of more knowledge, precision and ability. 

The proof: a few years ago there was a computer project that tested the "absolute" quality of moves, and it found a strong correlation between a person's chess rating (even over eras, for example 1970s compared to 2000s) and the absolute, objective strength of their moves. I think that the guy's name (who developed it) was Regan. 

I never bought the theory of "rating inflation"...

The result of writing first and reading later.... indeed, I was more or less repeating #3 above, and the guy's name WAS Regan. 

By the way - until about 1990, the FIDE rating floor was 2200. After 1990 and for several years it became 2000, and then "the floor opened up" and today we have 1000 ratings as well. If anything, this should have caused the ratings to deflate, not to inflate, due to (potentially) underrated players (such as children etc.) "draining" rating points out of the system...

We are where are. The best who did not have engines (Fischer, Karpov, Kasparov etc) should not be compared with those who do. I will always remember the excitement British Master Michael Stean felt when he announced (as Korchnoi's second) we have all of Karpov's games.
Today, not only is that easy but engines can be employed in study.
It seems to me that the current say best ten are like the best ten in the 1960's but with more generous or maybe more accurate grade ratings.

Dear Solskytz,

Rating inflation is "an increase in the average rating over all players in the rating system". This cannot be caused by a general increase in chess strength, as for each rating point won, one is lost. The Elo system only measures differences between players, it has nothing to do with absolute player strength. In a rating system made up only of beginners, where some were somewhat better than the others, you would still have players rated in the 2800s, and the average rating would be similar as in today's system (as long as entry rules were the same).

A different thing is an increase in the rating of a specific subgroup, for example, the top 10 or top 100. An increase in this subgroup can be caused by different things, such as a higher total number of players in the system (thus widening the distribution at the extremes of the bell curve), or an increase in the standard deviation of the distribution (caused, for example, by greater professionalisation, where some professional players would dedicate themselves much more to the game and thus be much better than the average player).