Yesterday, there was a lot of discussion about Grandmasters versus Super Grandmasters. The key problem is the rating inflation. Therefore, should the specification for the Grandmaster title be changed or raised to compensate for the rating inflation?
If yes, then what is your suggestion? If no, then could you please explain why not?
Thanks for engaging in this very interesting issue.
Chess Daily News from Susan Polgar
there is no rating inflation. why do chess fans always feel the need to “fix things” (draws, titles, etc)?
When will Afromeev’s dog be awarded the GM title?
Is 2550 and above for GM, and 2450 and above a reasonable standard for IM?
Re: commenter #1 saying “there is no rating inflation”, has anyone done a study of this?
The main question for me is: are there many more internationally-active players today? Are there more FIDE-rated games processed each year?
If so, then one would expect to have more players at all levels including the top. And one would also expect to have a higher overall peak (e.g. players between 2800 and 2850) just by “laws of large numbers…”
I’ve seen statistical studies instead which argue that the natural tendency of the rating system is to deflate—sorry, no time to find and link them now, just going out the door.
My subjective impression is that top player skills have improved especially in defending inferior positions, though this maybe hasn’t cut down the % of wins all that much. An analogy might be that in baseball, middle-relievers used when a team is behind are better than before, so there are fewer 10-run blowouts and teams lose 7-4 instead. My sense is that these 3 factors () greater population, () slight upgrade in skill, versus () natural deflation, have balanced each other out.
Of course, as I said in my comment to Susan’s other post, I’m actively working on objective measures to test arguments like the ones here, also intending application to other decision-making activities besides chess.
In the beginning the Grandmaster title was awarded to 5 players. For most of the 20th century the “grand” part of that represented the cream of the world’s chess players, but no longer. It’s time to create Super GMs with minimum ratings of 2750. The current list shows 8 players at that level, and there are a few who used to be as well.
Why do we need a rating deflation to justify the super-GM jargon? The GM title has never been associated with playing perfection.
A regular GM is (was and will be) a very strong, professional-level, player, someone who play stronger than chess masters. But there are players that are much stronger than regular GMs. What is wrong in recognizing them? Afterall, other sports use the words “star” or “super-star” in recognition of outstanding or extraordinary skills.
Creating new titles won’t really change anything. Players at 2700 are looking for a shot at the world championship not for another title.
The GM is a recognition to a life of chess, that’s why is permanent, the next recognition is World Champion. I don’t see any reason for something in between.
Great blog and an interesting article. Chess is such a great sport. I consider Gary Kasparov to be the greatest chess player ever http://hubpages.com/hub/Gary-Kasparov
It is silly to use a rating cut-off for “super GM” (2750 as some suggest). If rating inflation is indeed a problem then the Super-GM title will get watered down just like the GM title. Instead the super GM title should be based on making the worlds top 5 list. I haven’t looked but we wouldn’t have too many active SGM’s, just enough to make it interesting.
Zsuzsa,there isn’t thing like a chess rating inflation.More and more players reach 2600 elo mark simply becouse there are more and more players who objectively play at this level.It’s math ,nothing else.Why can they reach more easily such level of play today than before 20 years conclude yourself.You’re a smart lady:)
Hello
You have got a nice blog. There is a problem though. I couldn’t find any way to search your archive. It’s annoying to manually go through several months. Please add a search box to the layout of your blog. Here is a link to a ordinary search box that you can add to your Blogger layout.
How To Add Search Box
Hi,
GM is the title related to the level of play not with the place in the rating. There is being more people with that level of play then before.
It is simple.
Nothing to do is neccessary.
Rgds
Pony
“reach 2600 elo mark simply becouse there are more and more players who objectively play at this level”
There’s no such thing as an objective 2600 level of play.
The rating is not an absolute number describing the chess strengh. The rating is a relative number comparing a group of people.
When players at either end of the list, quit playing then there’s inflation or deflation. When groups of players only play amongst themselves then their ratings also lose correlation to the rest of the list.
According to the Wikipedia article “Grandmaster (chess)” there were, in 1972, 88 GMs. Today there are over 900. Various reasons are given for this, but the main reason cited is ratings inflation.
I accept that quality of play may have gotten a bit better over time, but I don’t think that this is what accounts for the inflation. The article mentions, for instance, the case of Nigel Short who, in the early 90’s, had a rating of 2650, which made him the third best in the world. Today 2650 would barely crack the top 100. Yes, players may have gotten better, but I don’t think they’ve gotten THAT much better in so short of time.
So, clearly, to me, the GM title has been devalued.
What to do about it?
I propose changing the standard for achieving a GM norm by raising the minimum performance for a norm by 100 points. I would also change the required rating to achieve GM-ness to 2600. To be fair to up-and-coming players, we shouldn’t make the jump to the higher standard all at once. Change the standard by 10 points per year over ten years. And potential great player, who is increasing in skill, can catch up and overtake this trend.
Along with this, I would append “emeritus” status to GMs who, while they have achieved the “lifetime GM” status, have fallen behind in terms of ratings. This status, only slightly devalued from “full GM”, acknowledges the fact that the player has achieved GM status at one time- but is no longer at the top of their game. (In what other sport do formerly great players get to hold on to their assumed status once their powers have failed?)
Brad Hoehne
Look to the left, you can search the blog. Find the Large icon labled “SEARCH THIS BLOG”
The problem is defining a standard metric. Talk of inflation will keep happening until we have a standard metric. In the past, it was not possible to have a standard metric since people have a limited life span.
However, today we could definitely have a standard metric since we have chess playing computers that have an infinite potential life span. A computer program does not die.
If FIDE wants to get serious about defining a standard metric they have the tools at there disposal. Simply take a pool of the top chess playing computer programs in the world at a certain date. Then you evaluate humans based upon how well they do against these computers.
Simple and it would stop talk of rating inflation.
GM: 2600, as it is today.
The problem is not in the ratings, it’s in the acquisition of the GM norms. I would propose a more difficult path to obtain GM norms than is currently in place. The difference historically is *not* player skill, but the increased availability of norm-qualifying events (NQE’s).
Since the number of NQE’s is much higher and more accessible, there should be an increase in the number of required norms.
Manyoso (two comments above as I wrote): In a sense that’s what we’re doing. But there just haven’t been enough human vs. computer encounters to rely on. And even if there were—if a human has the better side of a draw as Black all game and then falls into mate-in-one, what’s learned from rating the result?
Instead what we’re doing is rating the moves, by programs run to fixed high depths as a standard metric. However, the math behind doing so is far from simple! But you’re right—the idea is to stop “just talk” of rating inflation… And hopefully also to help stop cheating, as this work will establish “prior probabilities” needed to make thorough statistical tests in place of bandied-about impressions and unfounded allegations.
The root problem is that the purpose of the title GM has become confused. The original intent was as an honorific. It was bestowed to honor the five top players who did very well in the St. Petersburg tourney and few could quibble that these players were among the leading chess masters of the era (Marshall was probably the weakest, but there was nice geo-political balance – Capa, Lasker, Alekhine, Tarrasch, and Marshall – all of the major chess playing powers as it was were represented). But now it is not an honorific but a title of accomplishing some professional achievement – winning enough norms and keeping a certain rating.
When the title was honorific (remember it was not really formalized until fairly late in the history of chess) it was subjective but few dared to bestow it without really good cause. Who would call themselves GM in that era without a major tourney win over the major players of the era? Winning several 9 round once a month tourney isn’t the same thing as winning Baden Baden 1925 or London ’22 etc.
Here is a proposal:
If it is a title to designate professional accomplishment and fitness like a PHD or MD, then it should administered more objectively. All GM should be able to win a match against a suitably crippled computer program – like Fritz or Rybka running with limitations AND also show technical prowess – can they win an endgame regarded as as “WIN” and hold an endgame proven to be a “DRAW”.
“I would propose a more difficult path to obtain GM norms than is currently in place.”
I’d propose they strip you of your title, if you have, you jelaous blogger, you.
“Therefore, should the specification for the Grandmaster title be changed or raised to compensate for the rating inflation?”
None of the above!! You have no idea! Sorry!
Grandmasters have to understand that the ones who got their norms by unfair means are not forgotten. All will surface. Same goes for IM cheaters. Cheers to the honest players!
“I propose changing the standard for achieving a GM norm by raising the minimum performance for a norm by 100 points.”
I propose you stop making it difficult for others, just for you own blogger’s fun sake. You are no GM, nor will ever be, so please stop making our difficult chess lives even more difficult. Cool down, man!!
Obviously the problem is in fake titled players, or cheaters? We have an abundance of those.
What have you done in your chess carreer to prevent fake players surfacing?
“there is no rating inflation. why do chess fans always feel the need to “fix things” (draws, titles, etc)?”
Anon is completely right!
I agree with the posters above!!! There IS NO RATING INFLATION !!!
You need to know MORE to get a specific rating now than before. Players are BETTER!!!!
Thi sis common is almost every sport. Standards INCREASE.
I think that at lower levels (maybe under 2200) there is strong deflation due to FIDE reducing the rating floor. This makes it expensive to lose to a maybe under-rated 1800, and makes it essential for the 2200 to basically beat under-2000 all the time, which is not easy, to avoid losing points.
Also many of the under-2000’s are fast-improving and underrated juniors, even under-10 and such like.
Those who say there is no rating inflation have no idea what they are talking about and completely misinterpret what ratings represent.
They obviously paid no attention to anonymous 3:00 PM who correctly said [capitals mine]:
The rating is NOT an absolute number describing the chess strength. The rating IS a RELATIVE number comparing a group of people.
Thus for example, if we compare the relative dominance of the top player, we have Anand at 2803 today, and Fischer at 2785 in 1972. So is Anand 18 points more dominant over the rest today than Fischer was in 1972? No way!! Fischer’s relative dominance in 1972 would be something like 2900 in today’s terms.
In other words, there certainly has been inflation, and serious inflation at that.
Those who think there is no inflation should read Jeff Sonas’ attempt to deal with it on Chessmetrics, and Ron Edwards’ in his article at
members.shaw.ca/redwards1/
Edwards gives a pretty convincing case that inflation started seriously in 1986. Before that, the average rating of players ranked 11th to 50th stayed about constant for 15 years from 1971 to 1986. Then in 1986, it took off sharply.
I think one reason this might have happened is that in 1986, FIDE, for political reasons, increased all women’s except Susan Polgar’s ratings by 100 points. [This was an underhand agreement between Campomanes and the Soviets to keep a Soviet at the top of the women’s list, a position that Susan would otherwise have taken. Campomanes in return got the eastern bloc vote for his reelection in 1986.]
As Edwards says: “The sharp onset of the rating increase would also seem to be strong evidence against the idea that play is simply improving gradually over the years”.
What to do about it?
I agree with Brad Hoehne’s post.
The standard for achieving a GM norm should be raised by a hundred points.
Then I would take Edwards’ idea of the average of players 11 to 50 and adjust the standard by the change in that.
Playing standards at the top have, of course, increased, what with computer analysis and training and whatnot. And together with more players playing, there is a wider range in strength from beginner to top players than there used to be. Therefore, the introduction of a new Title, call it SuperGM or whatever, is certainly justified.
As for the comment ‘why do chess fans always feel the need to “fix things” (draws, titles, etc)?’:
Simple. They are broken. Broken things need fixing.
Odd things about ratings inflation. Take Susan herself for example. I think her FIDE rating is near 2600, but she hasn’t played regularly in nearly 10 years…the olympiad in 2004 and a few other matches with Karpov. Perhaps there a few more, but her rating is probably no longer accurate. Her actual strength may actually be 20-35 points higher than her current FIDE rating if she were to play again competitively.
I think Gata Kamsky was in a similar position a few years ago.
Blogger KWRegan said…
Re: commenter #1 saying “there is no rating inflation”, has anyone done a study of this?
As a matter of fact someone has:
http://members.shaw.ca/redwards1/
Thanks. The study is insufficient. It does not address the issue of the larger sample size today. The proposal to “rectify” ratings by “subtracting out inflation” looks like assuming what it’s trying to prove.