I'm trying to make sure I understand what SF means by slight deflation with a concrete example.
SF's Fischer example says that Fischer played at 2785 in 1972 and that would be the same strength as a 2785 player today.
I'm impressed they or at least Dr. Regan seemed to have made intrinsic rating work. I thought that it was possible, but still off in the future.
Dr. Regan says that Lasker, Capablanca, and Alekhine at their peaks were playing 2700+ chess by today's standards.
Regan says the big difference today is that there is not so much fall-off from the giants like L, C, and A to their less distinguished competitors.
He also says Capablanca was playing 2950 chess at New York 1927.
Why does the whole thing about inflation/deflation (w/e u guys r talkin bout) matter if its only 25 points? lol. doesnt seem like it would be significant for any good reason.
It matters because it has become commonplace even among some GMs to decry rating inflation. Ten years ago, the 2700+ club had two dozen members. Now, it is double that.
Ratings are higher than ever. Regan's data suggests that higher ratings are because players are stronger.
It should not be surprising that top chess players are better today. Football (American and global [soccer]) players are faster, stronger, and better skilled. Runners are faster.
Using intrinsic ratings, they should then be able to map the high end of USCF ratings to FIDE ratings.
That would be interesting.
Really fascinating stuff.
So unless there are flaws in Regan's approach or for some reason it can't work below 2000, it seems only a matter of time before intrinsic ratings are extended to class players.
In that case players could measure their skill levels without playing in certified tournaments.
One thing to remember is that in Lasker's day Elo's system wasn't used. Lasker used a colloquial sense of "master". I'm sure that I would be considered a master in his day because I won several tournaments and city championships and I teach chess. I'm not a master. When I was close, I played better than I do now. Ratings really do reflect performance, if not knowledge.
Cookie: You're one of my favorite commenters here, so if I've put you off, I didn't mean to.
There is software to demonstrate, assuming Dr. Ken Regan has got it right, which I don't take for granted though I find his work tremendously exciting.
The problem with ratings is that they are magic numbers which produce statistically significant rankings within a pool over large samples, but are less reliable with the small samples of tournament play, don't necessarily gibe with earlier histories of the pool, much less other pools, as well as real chess skill, whatever that is.
The USCF (and probably FIDE) tweak their rating algorithm now and then to get the ratings to better match some idealistic (or possibly political) goal for the ratings.
I've written chess ratings software and tried it out on game databases. I was somewhat surprised to see how much you could effect the ratings by minutely tweaking the algorithm parameters. Over large enough game samples those surprises go away, but in the fluid, somewhat random, human world of chess tournaments those samples aren't all that large and you can get significant undesired effects on player ratings.
Smyslov: Thanks for the Regan tip. I looked into him last year or so but it didn't register with me at the time.
As I understand Regan, we can assign a theoretical FIDE rating to Lasker and compare him on that basis with 2700+ players today.
According to Regan, Lasker, Capa and Alekhine would hold their own today, though not all the way to the world championship, unless Capa could have pulled out his NY 1927 peak game in which case he would probably win.
The Doctor should put Magnus in the TARDIS and take him to New York, 1927. That would be interesting.
Cookie: I added the 25 point thing. It was an arbitrary number just to be sure I had my plus and minus sign in line with what Smyslov was saying about a "slight deflation."
If you believe Dr. Regan, he has a method which can distinguish between 25 points with error bars of some sort.
You're making a classic mistake by taking a player's best result and treating it as his norm. There have been even better performances than Capa's in 1927.
I'm on my phone now so I can't look it up, but I believe Lasker's average IPR while he was world champion was around 2500.
In case I've buried the lede here, i.e. left the important bit insufficiently emphasized, Dr. Regan has a cpu-intensive method for calculating objective chess ratings, which he calls IPR, "Intrinsic Performance Rating."
Dr. Regan claims to be able to compare Capablanca with Magnus Carlsen and say that Capablance would have given Carlsen a serious challenge, and if Capablanca had played at his peak -- 100 points over Carlsen's average -- Capa probably would have beaten Carlsen.
That's a fascinating, compelling claim.
I don't know if it's true, but Regan is on the right track. He's comparing the output of top chess engines with human play and calculating a FIDE rating on that basis. If Regan doesn't get this precisely right, someone else will.
You're making a classic mistake by taking a player's best result and treating it as his norm. There have been even better performances than Capa's in 1927.
Smyslov: Geez, I compliment you and you kick me in the teeth. Duly noted.
I clearly qualified my statement:
"unless Capa could have pulled out his NY 1927 peak game in which case he would probably win."
I don't assume that he could, but if he did...
I'm on my phone now so I can't look it up, but I believe Lasker's average IPR while he was world champion was around 2500.
Smyslov: Lasker was one of the longest reigining world champions with most of his competition buried in the dust, so I'm not terribly surprised his overall IPR was 2500-something.
Nonetheless I will take Kramnik's assessment that Lasker was the "first 2700 player" over anything you will say in this discussion.
Carlson is MUCH better than Capa. Take a look at Capa's technique against Alekhine in slow chess. And compare it to Carlsen's technique in rapid chess this week. Carlsen's technique, especially against Ivanchuk was jaw-dropping!
Man, I hate autocorrect. I typed Carlsen, not Carlson. And please forgive my grammatical errors. The phone changed what I wrote.
Smyslov: Again I said, Capblanca at his peak in NY 1927 -- rating 2950.
Unless you want to discard Regan's methodology, that means that Capablanca at his peak was 100 points over Carlsen at his norm.
I said, "If Capablanca played at his peak game" he would likely have beaten Carlsen at his norm.
If you want to keep nitpicking a straw man, be my guest, but I made my qualification clear.
You are making a mistake. You are comparing Capa's best performance with Carlsen's average performance.
You are making a mistake. You are comparing Capa's best performance with Carlsen's average performance.
Geez. I get it. And I've made it clear. It was a hypothetical. Google that if you are unclear.
I found it interesting that Capa at his peak measured by Regan's software would be an off-the-chart 2950.
It would be interesting to hear Carlsen's peak performance is so we could compare apples and apples.
Why does the whole thing about inflation/deflation (w/e u guys r talkin bout) matter if its only 25 points? lol. doesnt seem like it would be significant for any good reason.