Reasons to believe there's FIDE rating "inflation"?

[Graham Price]

Re: Reasons to believe there's FIDE rating "inflation"?

The way I see the problem of ever increasing ratings is lost relevance.
It is very possible that a 2700 today is playing better than say, Karpov in the '80s - he (not " or she" yet) knows more opening theory, more endgame techniques, etc.

.

Ranked #1 Most times

Polgar, Judit (GM) (HUN) (61 times)



Highest rating

Polgar, Judit (GM) (HUN) - 2735
Top 50 Women July 2000

?not or she????

[Kenneth Regan]

Re: Reasons to believe there's FIDE rating "inflation"?

Now, to give you nightmares :p

Database errors will obviously affect the calcuated IPRs of the players -- garbage in - garbage out.

Disturbingly, Chessbase had (has?) errors even in games from World Championship matches: in their 2009 database I found a series of blunders at move 32-36 from the Botvinnik - Smyslov 1957 match game 9 (see below).

Nightmares indeed! My "Exhibit A" for this problem has been the game Sebag-Shen from the 2010 Ulaanbaatar Women's Grand Prix, where the omission of a Rook exchange at move 34 caused an insane number of queen-hang and mate-hang errors in a gamescore miraculously legal thru Move 57. Chessgames.com fixed it last year, but Chessbase Big 2013 database did not; you can find other examples I raised that are pending at Chessgames.com by searching for my kibitz-posts (KWRegan).

My data-collation scripts flag blunders. I have manually checked only the 6,000+ games in my main training sets for the 2200 thru 2700 levels in the reference years 2006--09, 1991--94, and 1976--79. I found 56 cases that struck me as likely erroneous, of which I felt able to fix 35 and "mitigate" another 10 or so; the rest I have no clue about. Extrapolation tells me there's almost a 1% error rate, and it inflates my unchecked error stats (and hence deflates IPR's) by about 10%. This I think contributes to why my data for games by players rated 1600 thru 2100, which I have not fully vetted (and it's harder to tell apart from "normal" blunders anyway) all fall below the projected line from my 2200--2700 data.

Alas I do have the Botvinnik-Smyslov error you cite uncorrected in my records, but as it is not in the training sets used to determine the IPR regression I have not yet vetted it. The new data format accompanying the current upgrade to the model which I mentioned will hopefully help some, and will be made public. In 1985 I manually edited the 24 x 18 dot-matrix grids of 12 fonts at 128 chars/font, for a system used for doctoral theses at Oxford University---that's 663,552 dots edited by hand over 6 weeks while recovering from a breakup. But as a happily married man with kids, for fixing gamescores this time I'll ask for help :D ---a panel commissioned by FIDE should be responsible for the "restoration" procedures anyway.

[Brad Thomson]

Re: Reasons to believe there's FIDE rating "inflation"?

If FIDE ratings are neither inflated nor deflated, does this necessarily mean that anyone today with a higher rating than Fischer's highest ever rating is therefore a better player than Fischer ever was?

[Felix Dumont]

Re : Re: Reasons to believe there's FIDE rating "inflation"?

If FIDE ratings are neither inflated nor deflated, does this necessarily mean that anyone today with a higher rating than Fischer's highest ever rating is therefore a better player than Fischer ever was?

I'd say yes. He was a great player in his time, and was definitely better than his opponents, but he definitely made more mistake than the actual top players (mostly thanks to computers).

[John Upper]

Re: Re : Re: Reasons to believe there's FIDE rating "inflation"?

I'm not sure what to say about Fischer, in part because he was such an outlier in his time, but also because his results the year before he quit were so good that his rating was climbing fast -- it dropped in his WC match vs Spassky, but not much -- and I'd guess that if he'd kept playing his rating would probably have gone higher. That is, I think Fischer's peak rating still lagged behind the actual quality of his play.

In principle, I think that if Fischer's rating had stabilized at 2780 then he'd have been playing worse chess than contemporary players with the same rating.

[Paul Bonham]

Re: Reasons to believe there's FIDE rating "inflation"?

I regularly hear people say that that FIDE ratings are inflated, but I've never heard a reason to believe it.

...

Here's my request:

post your best argument for the claim that today's players with FIDE-rating of nnnn (whatever) are not as good as the players with that same FIDE rating from the past.

If FIDE ratings are neither inflated nor deflated, does this necessarily mean that anyone today with a higher rating than Fischer's highest ever rating is therefore a better player than Fischer ever was?

I'd say yes. He was a great player in his time, and was definitely better than his opponents, but he definitely made more mistake than the actual top players (mostly thanks to computers).

All this talk about who was the greatest player ever is akin to talking about who the greatest player was who never made it over 2400. Here, I'll help get that started: Was Paul Beckwith the greatest sub-2400 player ever? Or Paul Morphy? (Someone in this thread mentioned that Morphy's highest rating in today's terms was 2344).

It's all a little silly because we have much greater players now who are head and shoulders above any human player. The search for chess perfection or the nearest thing to it should lead to the games of Houdini or Shredder or Stockfish or a dozen other chess engines.

I imagine someone's going to jump in with "That's like saying we shouldn't discuss who the greatest 100-metre runner ever was because there are cars that can outrace them." The reason we don't race cars against humans is because the race itself isn't something we analyze. We only tabulate the time it took to go a certain distance. But with chess, we analyze THE GAMES. We want to know where the mistake(s) occurred, and figure out what the best move was.

Despite this, chess engines aren't (usually) allowed to play in GM events, so we only have their games against each other, which are a gold mine for those who like to play through 100+ moves per side to find the exact moment where a tiny tiny sub-optimal move was made, and how that infitesimal imperfection was exploited flawlessly by the opponent to create a win. So what if it took 77 moves to force zugzwang? That's 77 beautiful moves (for some people anyways).

The problem in this respect is that chess is a game of perfect information, and as such, is only unsolved because of the size if its search tree. The rating accomplishments of the super-GMs means nothing viewed from this perspective. It's entertainment and nothing more. In fact, many GM vs. GM games of today and of the past may have existed for fractions of a second in the RAM of a computer playing one engine against another, and then thrown into the mythical bit bucket.

The discussion about rating inflation or deflation (given that it takes time spans of tens of years to become apparent) also means nothing. I would respond to the OP's request ("best argument for the claim that today's players with FIDE-rating of nnnn (whatever) are not as good as the players with that same FIDE rating from the past") with this: it's all just a waste of time. Would you actually take any time to determine who was better, Beckwith or Morphy?

I would suggest to any GM out there today, especially if making money from chess is not their priority, that if you want to become truly something remarkable in chess history, learn how to play chess in a manner such that you can have a plus score against the world's best chess engines. FIND A WAY TO BEAT THEM... REGULARLY. Don't even bother to play human GM's. Your goal is a plus score against all the computer engines, and then your rating will be well over 3300 and that will decide all these who's-better-than-who discussions.

Then when you've done that, do go ahead and play DOWN to the level of the super-GM's. Find out if your techniques can beat them. Would it be as easy as Mate Milinkovic playing down to the 1200 level? If not, and if in fact you cannot beat the super-GM's regularly and easily using your anti-computer techniques, I would say THAT is a major discovery and milestone in chess history. What exactly it would mean would be at least initially unclear and fascinating to discover.

[Mathieu Cloutier]

Re: Reasons to believe there's FIDE rating "inflation"?

I regularly hear people say that that FIDE ratings are inflated, but I've never heard a reason to believe it.

True, the ratings themselves might not be inflated.... but there sure is inflation going on. 50 players over 2700 (FIDE) is insane.

Not sure it's related to an increase in overall skill level. Of course the top 100 players today would beat the **** out of the top 100 from 50 years ago, but that doesn't make rating points appear out of thin air.

An extremely simplified example: 100 guys play against each other for years. The quality of their games will most probably go up, but the average rating will remain the same. Ratings won't magically go up because the guys play better moves. Rating points have to come from somewhere.

From wikipedia:

One possible cause for this inflation was the rating floor, which for a long time was at 2200, and if a player dropped below this they were stricken from the rating list. As a consequence, players at a skill level just below the floor would only be on the rating list if they were overrated, and this would cause them to feed points into the rating pool. In July 2000 the average rating of the top 100 was 2644. By July 2012 it had increased to 2703.

That may explain the ongoing inflation in FIDE ratings.

Also, maybe it follows with the fact that more people are playing high level chess? So you have more people at the end of the Bell curve? Then again, we would need data to prove this.

Anyway, as others have pointed out: we should not use ratings to compare players across different eras. In my mind, there's no way a guy like Akopian (currently 2700) compares to Tal (highest elo around 2700).

Mathieu

[Egidijus Zeromskis]

Re: Reasons to believe there's FIDE rating "inflation"?

The FIDE rating system had several players who were above all others by good margin: Fischer, Karpov, Kasparov, Carlsen. The first three are legends. Magnus still builds his legacy. The SCS (Secret Caissa Service) reported about his 2013 resolution - 2913 LOL


As for an inflation: it exists. Personal experience: a long time ago my rating was estimated ~2050 (thus it was not published in the list). After appr. two decades the established rating became ~2150. I don't feel that I'm playing better. I would say even worse LOL

Tal once answered that he was playing better chess compared to his 195x-6x attacks Though he is loved for those not-always-corrected sacrifices :p

[Kenneth Regan]

Re: Reasons to believe there's FIDE rating "inflation"?

From wikipedia:

Quote:
One possible cause for this inflation was the rating floor, which for a long time was at 2200, and if a player dropped below this they were stricken from the rating list. As a consequence, players at a skill level just below the floor would only be on the rating list if they were overrated, and this would cause them to feed points into the rating pool. In July 2000 the average rating of the top 100 was 2644. By July 2012 it had increased to 2703.

That may explain the ongoing inflation in FIDE ratings.

This is the kind of postulated knock-on effect my model applies to---and which my work RULES OUT. Well, back to academic-speak for a moment: my results show 16 points of inflation since the mid-1980's, with high confidence that it's between "nada" and "32". However, this strikes me as pretty much the difference I'd expect to see given the abolition of adjournments and faster time controls nowadays. Indeed most of the change is between the years 1990 and 2000.

Since unlike with cheating allegations no human dignity is at stake and there is no fear of suit, I'll come back from academic-speak to forum-speak with emphasis: the idea of a 150- or 100-point or even 60-point inflation bubble propagating all the way up to 2700 is false, falsch, falso, faux, fals, ложный, կեղծ, saxta, помилковий, שקר , كاذب, غلط, 假, 偽, 그릇된, vals, falsaj, ψευδής, झूठा, தவறான, hamis, rangar, bréagach, falsk, falskt, väärä, fałszywy, yanlış, sai... :p

Also, maybe it follows with the fact that more people are playing high level chess? So you have more people at the end of the Bell curve? Then again, we would need data to prove this.

Agreed.

Anyway, as others have pointed out: we should not use ratings to compare players across different eras. In my mind, there's no way a guy like Akopian (currently 2700) compares to Tal (highest elo around 2700).

Mathieu

It may take some in-depth probing of data, plus my model upgrade reflecting depth of play in various senses, but I suspect that Akopian is less tactically gifted but more accurate. You could also guess that Tal's health issues had an accumulated effect of 50--100 Elo; I would not be surprised to find that when he was "on fire" with the two longest lossless streaks in 1972--1974, he was playing 2800+. That would satisfy both your intuition (and mine) and what my data say.

[Hans Jung]

Re: Reasons to believe there's FIDE rating "inflation"?

Yes it is sure a different chess world recently. Just look at the number of tournaments offered all the time around the world and the easy access to a number of ways to improve. The biggest striking contrast is the length of time it took to see critical tournament games after they were played as to now when they are available in hours. 50+ 2700 players is not insane and it is continuing to grow that way. The chess world continues to mushroom explosively and is in no way comparable to the insular chess world of just 30 years ago.

[Garland Best]

Re: Reasons to believe there's FIDE rating "inflation"?

So you have more people at the end of the Bell curve? Then again, we would need data to prove this.

I believe that the main reason is that the rating system IS designed to be a bell curve, with a rating peak at 1600, and there are more people who are rated now.

For arguement's sake assume that 0.05% of rated players have ratings over 2700. If in the 80's there were 10,000 rated FIDE players, that would equal 5 players with ratings over 2700. If today there are 100,000 FIDE rated players, then we expect 50 players with ratings over 2700. That's how Bell curves work.

FIDE's regular lowering of the rating floor in their system complicates matters, but I recall an article on Chessbase showing how the ratings today still closely follow the same bell curve as 30 years ago. This does not indicate rating inflation.

Regardless, the other tenet of the rating system is that someone with a 200 ratings point advantage will get 3 points out of 4 in a 4 game match. Therefore one should look at the difference between the number 1 player and the number 2 player to see how much the best player dominates his opposition. The bigger that gap, the better that player is than the rest of the world. That is how I would gauge who the greatest player in history is.

[Roger Patterson]

Re: Reasons to believe there's FIDE rating "inflation"?

Regardless, the other tenet of the rating system is that someone with a 200 ratings point advantage will get 3 points out of 4 in a 4 game match. .....

Although that is the theoretical basis underlying the rating system, in actual fact, in neither the FIDE rating system or the CFC system do actual results as a function of rating difference match the expected score. The actual vs theoretical for the CFC is plotted at http://www.victoriachess.com/cfc/opponents2.php. Jeff Sonas has plotted results for the FIDE system in several articles (one of which is referenced at the above link).

Why this is, I don't know. But, it does potentially have inflationary consequences (the actual curve has lower rated players scoreing better than predicted, if new players are typically weaker than the people they play, then their provisional starting rating will be systematically too high). Haven't looked too carefully at how big a factor this is.

[Mathieu Cloutier]

Re: Reasons to believe there's FIDE rating "inflation"?

But, it does potentially have inflationary consequences (the actual curve has lower rated players scoreing better than predicted, if new players are typically weaker than the people they play, then their provisional starting rating will be systematically too high). Haven't looked too carefully at how big a factor this is.

Good point! This, combined with rating floor may be the recipe for an inflow of rating points in the system.

Mathieu

[John Upper]

Re: Reasons to believe there's FIDE rating "inflation"?

(the actual curve has lower rated players scoreing better than predicted...

could that be because lower-rated players who have just recently got their first ratings tend to be the quickest improving ?

[Patrick Kirby]

Re: Reasons to believe there's FIDE rating "inflation"?

I tend to agree with Tom's point of view is just an estimate of strength relative to a certain pool. a specific rating can be inflated or deflated relative to other players playing in the same pool, but ratings as a whole can't be deflated or inflated since a rating is by definition a relative not absolute metric.
For the same reason there's no basis to compare the ratings of players from different eras since they competed against different pools of opposition.