Rating Inflation

k

For a time, when Kasparov was 2800+, Karpov was the only 2700+. Today there are 42 2700+ players. Do you really believe there are 42 current chess players as good as Karpov was then?

Wikipedia has a good section on ratings inflation/deflation.

The basic premise with ratings inflation is that a player of a specific rating today, say 2400, is weaker than a 2400 from a previous generation. So, in your example of more 2700's, the theory is that in the past, those players would have been rated less so their ratings are inflated over their true strength.

That said, ratings are really only useful in the context of expected outcomes of a game between two players. A specific rating system should try to, as accurately as possible, be able to predict the results between ratings over a number of games.

In USCF, at least, there is also a rating floor.  Someone who's attained 2200 points, for example, might have a floor of 2000 points.  So, if they end up actually performing at a level below 2000 once they reach old age or become out of practice, they'll still retain their 2000 points and they'll still award points to their opponents based on 2000 point playing levels, rather than the 1900 level or whatever that they're actually performing at.  In the end, that all means that slowly but surely those older players who've become weaker will donate rating points that didn't otherwise exist.

Let's put it this way: would you compare a chess.com 2700 with Karpov?

Ratings are relative, not absolute.

Ratings are a statistic distribution... a relative measure.  You're thinking of them as an absolute measure as if taking a measuring tape to a person's skill.  Ratings say my skill is in ____ proportion to the group.  Not my skill is _____ cubic meters.

So if the group is weak, your rating will be high (if you played against nothing but new players you may be rated 2300).  Or if the group is very strong, your rating will stay low (if the top 10 players started with new ratings, and played noone but eachother they would all be rated below 1300).

The ratings jumped very fast in a short time period, there are several explanations that try to account for this, but no one knows for sure.  For example, the group may have changed by injecting more new players.

Exactly. FIDE ratings are harder to get because you have to beat some really good people to get those ratings, they have a lot of excellent players to beat you down. Chess.com has relatively few titled players and experts, most people are here to learn or play casual chess, not to win elite tournaments; this leads to less pressure from above, and your rating soars (assuming you are good to begin with).

Likewise, in the days of Kasparov and Karpov, you had those two giants beating down the others, yet even the ones they beat were really top-notch. Nowadays you have a large cadre of people drawing to each other at the top (which prevents anybody from losing many points), while they eat up the smaller fry and binge their ratings. That is how there can be so many high-rated players without an increase in actual talent.

DZumpino94 the question of if there is rating inflation is academically interesting, but as a practical matter the reality is that the rating system is designed to only be applicable within a particular context. There is no reason to expect that one can compare ratings from different time periods anymore than one can compare ratings from any other seperated rating pools. 

Not quite. Imagine if some really talented guy came out of nowhere and won game after game against Anand, Carlsen and all the others. Their ratings would go down obviously, because they are losing to this new guy. So now they have lower ratings. But they are not less talented than they were before. Their ratings have adjusted to a change in the pool of players, not to the strength of the players themselves.

Drawing doesn't hurt your rating if it is against someone at the same rating. The important point is that these guys are not losing many games.

Yes, ratings still represent player's strength, but only compared to the other players. And, much like the real value of money, the way that ratings are affected can sometimes be strange and non-intuitive.

lies, damned lies, and statistics.

I love your assessment. I think it is easier for people to have something to compare things with. I like the idea of comparing to weights and measures to chess ratings. People are used to a set standard that all else is measured by. When they try to adapt their thinking, to a standard in constant flux, suddenly they are bewildered, angrily feel cheated and struggle to find a quanitification for it.

I disagree.  I think it's comparing the rating system to things it's not (such as currency by lordnazgul) that gets people confused in the first place.  Ratings measure a probability of winning based on past performance... it was never a measure of anything absolute like strength... that's the main problem the OP is having in the first place.

This can't be true. They would have have to understand the rating system first in order for it to be so. Besides, not everyone equates things the same way, before coming to the proper conclusion.

As far as the rest of your statement. It has its merits.

P.S.  F.Y.I.   ...Without anything to compare something to, we wouldn't know the differences between ideas. We need comparisons for proper perspective. This is how we establish standards.

(EX.) Someone chose a "particular" length once. Everything else that was measured by it and measured up to it, become known as, what we call an inch.

P.S.S. There is no quantification without a way of comparison.

There's absolutely no evidence or logic to support the believe that ratings represent absolute strength. I don't understand why it's so hard for people to understand the concept of relative versus absolute, unless they do and they actually think 2800 rating points represenst some exact measure of a person's ability, which is what they seem to be saying.

Several people here have explained the concept very well I think and if you don't understand you ought to re-read the posts very carefully.

If you want to teach an idea, then you're absolutely right, IMO you have to start with something the person already knows and either build on it or relate the new information to it.

If you want to resolve a confusion about something a person has learned incorrectly, then ideally you're as specific as you can be about that idea.  For example it wont help to say ratings are a lot like a dollar which represents a value... the person will wonder why doesn't a rating represent a strength?  I think if they're up for it, it would great for them to work though the derivation of the formula on Wiki to understand where those numbers come from in the first place.  I can tell you as input they don't take into account things like your ability to calculate, the endgames you know, or even how you preform under pressure :)

However (IMO) it doesn't take a background in math to understand the concept.  Not to be rude, but maybe all it takes is a little work on the asker's part such as looking up the words probability, relative measure or even standard deviation.

As a side note, IMO the inflation shouldn't be hard to figure out really.  If you gave a statistician a computer, all the rating data, and a few weeks it would be resolved without question.  (Unless I'm underestimating the hidden variables that mess this stuff up).

Karpov in his prime would beat any of the 2700 players around now. 

Karpov in his prime would beat Anand, and Arnonian, and Carlsen right now.

Unsupportable fanboi assertions such as these are what make discussions such as these so interesting. 

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.