Is rating deflation a thing?

Discuss anything you like about chess related matters in this forum.
Brian Valentine
Posts: 450
Joined: Fri Apr 03, 2009 1:30 pm

Re: Is rating deflation a thing?

Post by Brian Valentine » Mon Dec 23, 2019 7:24 pm

I’m sure that I will not be the only one interested in the results. I did start some work on this to see if something similar could be included in the monthly grading suite of tests. However it went into the too difficult pile, where it remains.

There are three principle hurdles. The main one is deciding on what measure to use for success (or relative failure). To correlate with results there has to be something that weights the critical points in a game higher than more placid situations. The second one is how to determine when the opening has finished and so the player can be fully evaluated (noting that there should be some credit for superior opening knowledge). The third is when to stop evaluating, and this is pertinent for Keith, based on his reputation at least. The computer evaluation is more wayward when the tablebase is measuring large swings, while players are down to increments. From what I have researched designers tend to cut the evaluation off just at the point the better player has the momentum in a grind.

As for monthly grading, the decision is to monitor movement in average grade and dispersion around that average. If these are maintained then grading limited events will attract similar percentiles of the playing population and this outcome would be better than what has happened in ECF grades over the last ten years.

User avatar
JustinHorton
Posts: 6918
Joined: Mon Aug 04, 2008 10:06 am
Location: Somewhere you're not

Re: Is rating deflation a thing?

Post by JustinHorton » Mon Dec 23, 2019 8:02 pm

Keith Arkell wrote:
Mon Dec 23, 2019 6:04 pm
JustinHorton wrote:
Mon Dec 23, 2019 3:24 pm
I'm guessing, by the way, that with the use of computers we could come to some kind of conclusion about the comparative strengths of players today and x years ago, but it'd involve the input and analysis of a very large number of games.
The site Chessdb does a weaker form of this, by taking a, presumably, random sample of all of our games and putting a percentage on our level of accuracy.
That's interesting, though what I was mostly thinking of was comparing the chess played by players rated n in year whatever with the chess played by players rated n in, say, 2019.
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com

Daniel Gormally
Posts: 56
Joined: Wed Dec 11, 2019 6:41 pm

Re: Is rating deflation a thing?

Post by Daniel Gormally » Mon Dec 23, 2019 8:12 pm

As far as explosion in the standards of players are concerned

I wouldn't say players are that much better than they were say 20 or 30 years ago, just that there are far more good players than there were. The reason for that is obvious- the internet. Which makes it much easier to become a strong player, and computers in general being stronger has an effect.

If I play in opens now (and sunway stiges seems a prime example) then the average age of the opponent is like, I dunno, early twenties? that seems significantly younger than say 2006, when I got the grandmaster title. So I can't say for sure but seems more difficult than it ever was to achieve a high rating, particularly if you are just playing in opens, as every other opponent is some Indian 14 year old rated about 2100 and deadly at tactics. And of course incredibly motivated and fearless.

I think because of computers standards have gone up, people know better than ever how to deal with certain positions, and the general level of defence has got higher, because computers have taught us that many positions that we thought were terrible, are in fact not that bad and in some cases better for us. But I think the most telling factor even more than "rising standards" is just that the strength in depth is far, far greater than it was, even 15 years ago.

Jonathan Rogers
Posts: 3935
Joined: Tue Nov 18, 2008 9:26 pm

Re: Is rating deflation a thing?

Post by Jonathan Rogers » Mon Dec 23, 2019 8:22 pm

Keith Arkell wrote:
Mon Dec 23, 2019 6:04 pm
JustinHorton wrote:
Mon Dec 23, 2019 3:24 pm
I'm guessing, by the way, that with the use of computers we could come to some kind of conclusion about the comparative strengths of players today and x years ago, but it'd involve the input and analysis of a very large number of games.
The site Chessdb does a weaker form of this, by taking a, presumably, random sample of all of our games and putting a percentage on our level of accuracy.

...
Random, I suppose, though in my case it takes only five games, four of them surely among my worst games where I nonetheless tended not to lose (including my well known draw with Keith a few years back when we both missed mate in two!)

Anyway.

I wonder whether we have been interpreting the word "stronger" differently in this thread. For Keith, it means correctness or accuracy of each move, so SF can play the major role. For Danny as well as me (I think) it basically means getting results against fellow humans, which may entail selecting surprising openings, playing moves that are known not to be best, or just ahving the habit of outlasting opponents in error-strewn games. And that may be one reason for the different opinions. We still don't think that Keith would beat his 1999 self, even if his accuracy has improved.

It would still be impressive if it had, though.

Daniel Gormally
Posts: 56
Joined: Wed Dec 11, 2019 6:41 pm

Re: Is rating deflation a thing?

Post by Daniel Gormally » Mon Dec 23, 2019 8:43 pm

One player who certainly could have argued that he was stronger in his late forties and early fifties was Korchnoi, who had his peak period around about then.

Although in his case I think it might have been more about changing circumstances than anything else, possibly he flourished being a big fish in a little pond when he was away from Russia, which wouldn't have been the case when he was just another strong grandmaster living in the Soviet union during the 1960s. in 1974 though, when he started his run and began to challenge for the world title, he was still living in Russia, so it could also have been the fact that he was in the right place at the right time, Fischer had dropped out of it and apart from Karpov there wasn't anyone else to be afraid of. And Korchnoi was a uniquely motivated individual, who kept that level of passion going much later into his life. I don't think most players in their forties and fifties have the desire of a Korchnoi (or of a Keith) to keep improving and get good results in chess.

In chess clearly your results don't depreciate as much as they would in sports like football or tennis, but then chess isn't really a sport. I do think there should be some depreciation though. In terms of memory, energy, ability to keep motivating yourself. It could be that Keith underestimates how strong he was in 1999. It could well be that getting out a marriage that he wasn't particularly happy in, freed him up and lead to him playing exceptionally well during that period. It's also easy to forget that chess sometimes is about form, and sometimes that form can last for just a short period. Maybe Keith's memory relates to that general period, where he felt he wasn't very strong, and that shorter period where he got to 2545 he forgot about and forgot about how well he was playing then?

Adam Ashton
Posts: 83
Joined: Wed Feb 24, 2010 7:37 pm

Re: Is rating deflation a thing?

Post by Adam Ashton » Mon Dec 23, 2019 9:39 pm

Jonathan Rogers wrote:
Mon Dec 23, 2019 8:22 pm
Keith Arkell wrote:
Mon Dec 23, 2019 6:04 pm
JustinHorton wrote:
Mon Dec 23, 2019 3:24 pm
I'm guessing, by the way, that with the use of computers we could come to some kind of conclusion about the comparative strengths of players today and x years ago, but it'd involve the input and analysis of a very large number of games.
The site Chessdb does a weaker form of this, by taking a, presumably, random sample of all of our games and putting a percentage on our level of accuracy.

...
Random, I suppose, though in my case it takes only five games, four of them surely among my worst games where I nonetheless tended not to lose (including my well known draw with Keith a few years back when we both missed mate in two!)

Anyway.

I wonder whether we have been interpreting the word "stronger" differently in this thread. For Keith, it means correctness or accuracy of each move, so SF can play the major role. For Danny as well as me (I think) it basically means getting results against fellow humans, which may entail selecting surprising openings, playing moves that are known not to be best, or just ahving the habit of outlasting opponents in error-strewn games. And that may be one reason for the different opinions. We still don't think that Keith would beat his 1999 self, even if his accuracy has improved.

It would still be impressive if it had, though.
I have a 'quality of play' rating of 87.83, so we can probably bin Chessdb as a reliable source...

Tim Spanton
Posts: 325
Joined: Thu May 01, 2008 11:35 am

Re: Is rating deflation a thing?

Post by Tim Spanton » Mon Dec 23, 2019 9:46 pm

Some others:
Spanton 72.35
Arkell 87.05
Gormally 82.44
Rogers 65.72
Carlsen 87.39

https://beauchess.blogspot.com/

Daniel Gormally
Posts: 56
Joined: Wed Dec 11, 2019 6:41 pm

Re: Is rating deflation a thing?

Post by Daniel Gormally » Mon Dec 23, 2019 10:00 pm

I think there's an obsession with accuracy now though. A lot of it stems from computers and all these websites like chessbomb. I said in a video I just filmed that accuracy is a bit of fallacy in chess, because the game is so complex. Look at the london classic fide open lots of horrible blunders going on, and this is supposed to be a reasonably high quality event.

Some players like carlsen have a very accurate style, they play that sort of way, someone like kasparov would score lower on the accuracy index, but it's just a different style of playing.

Adam Ashton
Posts: 83
Joined: Wed Feb 24, 2010 7:37 pm

Re: Is rating deflation a thing?

Post by Adam Ashton » Mon Dec 23, 2019 10:15 pm

Daniel Gormally wrote:
Mon Dec 23, 2019 10:00 pm
I think there's an obsession with accuracy now though. A lot of it stems from computers and all these websites like chessbomb. I said in a video I just filmed that accuracy is a bit of fallacy in chess, because the game is so complex. Look at the london classic fide open lots of horrible blunders going on, and this is supposed to be a reasonably high quality event.

Some players like carlsen have a very accurate style, they play that sort of way, someone like kasparov would score lower on the accuracy index, but it's just a different style of playing.
Yeah, I agree completely. 'Accuracy' is obviously easier in simple positions which is why positional players always top these kind of computer matching comparisons.

Keith Arkell
Posts: 745
Joined: Tue Dec 30, 2008 2:10 am

Re: Is rating deflation a thing?

Post by Keith Arkell » Tue Dec 24, 2019 1:08 am

Deleted ( posted twice)
Last edited by Keith Arkell on Tue Dec 24, 2019 10:55 am, edited 1 time in total.

Dragoljub Sudar
Posts: 124
Joined: Wed Jun 16, 2010 10:34 pm

Re: Is rating deflation a thing?

Post by Dragoljub Sudar » Tue Dec 24, 2019 1:28 am

Alex Holowczak wrote:
Mon Dec 23, 2019 10:23 am
Actually, if we submitted more of our chess for FIDE rating in England, then I think this problem would slowly disappear; so to some extent the same adults who don't want to FIDE-rate more sections are the ones who are helping to cause the problem.
The ECF can help solve this 'problem' by not forcing us to pay extra money to become Gold members in order to play FIDE rated chess.

Keith Arkell
Posts: 745
Joined: Tue Dec 30, 2008 2:10 am

Re: Is rating deflation a thing?

Post by Keith Arkell » Tue Dec 24, 2019 1:35 am

Keith Arkell wrote:
Tue Dec 24, 2019 1:08 am
Adam Ashton wrote:
Mon Dec 23, 2019 9:39 pm
Keith Arkell wrote:
Mon Dec 23, 2019 6:04 pm


The site Chessdb does a weaker form of this, by taking a, presumably, random sample of all of our games and putting a percentage on our level of accuracy.

I have a 'quality of play' rating of 87.83, so we can probably bin Chessdb as a reliable source...
From only 5 games though, which would be unreliable however accurate the method was. In Danny's case they subject 21 games to the test, I get 61 put through the mill, while Carlsen has about 1000 games dissected.

I do agree with Adam's point though that these measurements tend to favour positional players, so, for example, Capablanca scores very highly ( 89.28!) whereas Conquest - pretty much the opposite of a positional player - only scores 83.61
JustinHorton wrote:
Mon Dec 23, 2019 8:02 pm

That's interesting, though what I was mostly thinking of was comparing the chess played by players rated n in year whatever with the chess played by players rated n in, say, 2019.
I agree that this would be far more interesting, but I'm not sure where to look to find such work. I know that Professor Kenneth Regan has done a lot of impressive research cross-referencing engine assessments with human play, but this might be mostly related to cheating detection.

Regarding Adam's point that engines tend to be more impressed by the level of play of positional players because you get more Chessbomb style red moves in random, impossible to calculate positions, I don't think this would be an issue with the experiment Alex and myself are hoping to put into action - viz comparing the level of play of 1999 Arkell with that of 2019 Arkell. Why not? Because my style is pretty much the same now as it was back then!

I'm not sure I can agree with Jonathan's point that I might be more accurate today, but that would be negated by me somehow being less resourceful or less creative or less imaginative, and so on. I tend to be with Bobby Fischer on this one. Whatever the position you get it's about playing 'good moves'. I play as many interesting combinations today as I did 20 years ago - perhaps more, because psychologically I am more freed up / less inhibited.

Assuming we go ahead with this experiment ( and I very much hope that we do) if a 3600 rated engine assesses my play as being better today that it was 20 years ago ( and I'm confident that it will!) I will then consider the matter decided.

To argue that my higher level of accuracy today reduces my need to be resourceful in dubious positions, which is just as well because old players lose the ability to handle dubious positions, is just blather, and is no different to saying that if Karpov wasn't so strong positionally then he would have got outplayed tactically over and over again by John Nunn. The whole point of accurate chess, of strong positional chess , is that it transcends the game above the need to be tactically resourceful. Carlsen, Karpov, Ulf Andersson and Capablanca all demonstrate/d this time and time again.

Daniel Gormally
Posts: 56
Joined: Wed Dec 11, 2019 6:41 pm

Re: Is rating deflation a thing?

Post by Daniel Gormally » Tue Dec 24, 2019 9:07 am

"To argue that my higher level of accuracy today reduces my need to be resourceful in dubious positions,"

You're clearly ignoring your games in the Benoni then :D

User avatar
JustinHorton
Posts: 6918
Joined: Mon Aug 04, 2008 10:06 am
Location: Somewhere you're not

Re: Is rating deflation a thing?

Post by JustinHorton » Tue Dec 24, 2019 10:24 am

Keith Arkell wrote:
Tue Dec 24, 2019 1:08 am

JustinHorton wrote:
Mon Dec 23, 2019 8:02 pm

That's interesting, though what I was mostly thinking of was comparing the chess played by players rated n in year whatever with the chess played by players rated n in, say, 2019.
I agree that this would be far more interesting, but I'm not sure where to look to find such work.
I don't think there is any on any scale, although I seem to remember John Nunn doing some work some years ago to try and see how strong he thought the competitors in some old tournament were. (Can't recall the detail, maybe somebdy else can.) But it ought in principle to be possible.
Last edited by JustinHorton on Tue Dec 24, 2019 11:34 am, edited 1 time in total.
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com

Keith Arkell
Posts: 745
Joined: Tue Dec 30, 2008 2:10 am

Re: Is rating deflation a thing?

Post by Keith Arkell » Tue Dec 24, 2019 10:54 am

Daniel Gormally wrote:
Tue Dec 24, 2019 9:07 am
"To argue that my higher level of accuracy today reduces my need to be resourceful in dubious positions,"

You're clearly ignoring your games in the Benoni then :D
It wasn't me presenting the argument!

Post Reply