Rating Inflation

Im simply offering my opinion, and obviously i have no way of proving it.  But like i said, id take Karpov in his prime against any GM today.  And i rooted for Korchnoi in his matches against Karpov. 

 Is it so outageous an assertion when you consider that Karpov, long past his prime, had plus scores against many players still holding their own at the top (Kramnik, Topalov, Ivanchuk, Kamsky etc.) and also  beat Anand in a match?

To be clear, I have no problem with people pulling for thier favorites in discussions of who the best was or how they'd do today. I'm just noting that such assertions are baseless. Even if they're correct, they're not provable. 

Nor am I being entirely fascitious. If it wasn't for these sorts of sidebars, these discussions would be made up of 3 posts made by people who understand the subject, 4 folks asking followpup questions and a half-dozen folks who don't know what they don't know making false statements that need to be corrected. 

The reason rating inflation is a problem is precisely because questions like this are questions we really want answers to, but for which we'll never have a fully satisfactory response because we don't have a way to do anything but measure a person's relative ability within a particular moment of history, and not across time periods.

Karpov didn't destroy Adams when they faced each other, including all games Adams has a plus but in classical it is +2 -1 =6 to Karpov. All decisive games played before 1995, when Karpov was still quite strong and Adams far from a top ten player. He improved a lot later and was stable top 5 the first years of the 2000s.

Karpov was 2775 in 1996, and I think it's quite possible that Ivanchuk of today (2765) is playing chess on approximately the same level as Karpov did 15 years ago. Chess has developed quite a bit, and with shorter time controls and no adjournments it has at the same time become more difficult, but I think the inflation argument has been exaggerated.

Did the 2700 rated Gelfand from 21 years ago play better chess than the 2727 rated Gelfand of today? I have no idea, it's hard to compare. Fischer's 2785 as #1 was obviously more impressive that Anand's 2786 as #1 35 years later considering the distance to #2, but it's hard to say if Fischer's moves were stronger than Anand's, and if so, how much stronger.

I suspect that sheardp is touching upon the primary cause of ratings inflation.  As new players enter the pool of possible players, they're allocated some initial number of ratings points (e.g., 1200 in the case of chess.com).  These newly allocated points then become available for others to acquire when they beat these new players.  Those players, in turn, will then have a higher number of points than they would otherwise, which better players can then acquire when they beat those players... and so on, until it reaches the highest levels.  In other words, there's a general funneling effect where newly allocated ratings points eventually reach the highest levels.  Of course, the pool of available points is also depleted when people stop participating.  But if the total number of newly available points (from new participants) exceeds the number lost through depletion, the overall effect will naturally be one of inflation.  I suspect that the demographics of the number of new rated players over the last few decades would bear this out, and would correlate fairly strongly with indicators of ratings inflation (of course, the calculation of correlation would have to take into account some lag time for the newly allocated points to be re-distributed).   

I have always felt that there is too much obsession with rating. It is like grades in school. In the olden days the question of who was the better player was settled over the board rather than by comparing ratings. That's why they started having tournaments in the first place.

I intentionally didn't disclose my opinion about ratings inflation so that I could make a point without someone saying this.

The relativity of ratings allows for ratings inflation/deflation, but it doesn't provide any evidence that it does or does not exists.

We can see that ratings are not absolute representations of strength just by understanding the mathematical formulas that govern them. This is one reason I like mathematics. Opinions have no real place there.

It seems to me that it's really not possible to determine with certainty. I think the only certain way to do it would require the use of an objective evaluation of playing strength. The closest we have to this would be software making use of a modern chess engine, which in my opinion isn't good enough to judge top rated humans accurately. It seems to me that there are far too many variables to do it any other way.

Someone has already tried to do this and has called his method chessmetrics or something.

If all tournaments were unrated people would still see that Carlsen, Aronian and Kramnik had some good results lately, and that is also why they are top three on the rating list. It's all still settled over the board, ratings or not.

This may be a bit late in the discussion, but if the concept of rating as an absolute vs. a relative measure is still unclear, compare it to player ranking in, say, tennis. Djokovic is number one today, but there's no way of determining whether he's now good enough to beat Agassi when he was in his prime. This is the point that many people have tried to explain.

As for how this relates to chess ratings, think of it this way: tennis is a closed pool where there is a yearly defined amount of players. This allows their "rating" (or "seeding") to be discrete - i.e. everyone gets a unique slot in the ratings chart, which can then be mapped out by giving everyone a unique integer in ascending order from best to worst. Djokovic is #1, Nadal is #2, Federer is #3, etc. Now, chess is an open pool with a far greater number of players, and new ones can enter the pool at any time. So, you have to give these new players an approximate ranking (a "rating" of, say, 1200), and adjust it with each performance sample. The difference here is that since we want the top to be open-ended (to allow the order to change at any time without performance samples from each player), we invert the order of ranking and say that the higher the rating, the better the player. These rankings are not unique, since two or more people can have, say, 2231 as their rating. The ratings for the top players are now around 2800, and they are not consecutive; as sparse as they are amidst top players, they effectively translate to #1 (Carlsen at 2835), #2 (Aronian at 2805), #3 (Kramnik at 2801), etc. but they are viable to change at any time.

Note that in both cases, the rating is relative, not absolute. Just as we cannot know if Djokovic today could have beaten Agassi twenty years ago because both happened to be #1 at a given time, we cannot simply state that Carlsen could easily beat Fischer because he now has 50 rating points more than Fischer did at his prime forty years ago.

As for ratings inflation - a relative rating scale is obviously going to fluctuate as people with different skills enter and exit the rating pool. Say, if an undefeatable genius enters the pool and plays against people, his rating increases and that of others decreases. If he then exits the pool, the average rating of the remaining players will have decreased because of his presence. Statistically, this effect should be balanced out over time, but artificial disturbances can shift the average in one direction or another. For example, lowering the minimum rating from 2000 to 1200 allows more low-end players to enter and then exit the pool, which raises the average rating of the original pool of players if there is any interaction between them, thus causing relative "inflation" as compared to absolute skill levels.

But at the same time it's obvious that Djokovic would beat Lacoste in his prime since tennis has become a much more professional sport than it was in the 1920s, and that doesn't in any way reflect badly on Lacoste.

With a longer perspective it's clear that the best chess players today have a superior knowledge of not only opening theory and endgames than for example Anderssen and Steinitz in the 1860s, in spite of having less time for the moves and no adjournments. So even if Steinitz relatively speaking was greater than Aronian since he was the best player in the world for decades I don't think anyone doubts that Aronian's moves are stronger than those of Steinitz.

As for stating that Carlsen easily would beat Fischer because of those 50 points, I don't think anyone would say that Carlsen easily would beat Karjakin today because the latter is 70 points below. Anand is 2799 and maybe he would have been an even match for the 2785 Fischer since much has happened the last 40 years not only with regards to opening theory.

Well, yes, and surely that also applies to chess, but it still has nothing to do with their ratings. Just like you can't deduce whether the #1 tennis player today is better or worse than the #2 player from decades ago only from the fact that he was ranked higher, you can't use chess ratings to conclude that Carlsen, or even Kasparov, are better than world champions from half a century or more ago who had considerably lower numbers as their rating. And this is the misunderstanding that people in this thread have been trying to correct.