can magnus carlson cross the 3000 elo barrier

Better question: can Carlsen quit chess and go do ballet or something that better suits him?

"I am trying to eradicate ignorance about how the ELO system works."

Sure. I'm just saying I could agree with your general points about how it's easier to lose rating points when you're higher rated, and your point about his results this tournament, and still say this doesn't (in itself) cause much concern for him getting to 2900 (although again I would not call it trivial).

In other words it is possible to not be ignorant of your points and at the same time think his chances of getting to 2900 in his lifetime are good.

Well, it does depend on how many draws are in between those wins and losses.  The post WCC result you are referrring to is a tourney where Carlsen only had wins and draws, and never lost a game.  If he can do that til the next WCC this fall, he should make 2900, I think.  That calls for say a half dozen tournaments with the same opposition levels, 3-4 wins, and no losses.

He should do ballet, it makes more sense for a fashion model.

Ok, but I never said anything about him not being able to reach 2900.  I said somewhere in this thread that he would not reach 3000, though.  To do that he would need a boatload of 2850-ish players and a sprinkling of 2900s as peers, not 2750-ish with a couple 2800s.

All that would do is exchange winning more often with beating better players. Sure, Magnus would earn more points beating 2850s. But he wouldn't win as often and would lose more :)

Maybe for some reason he would enjoy playing better players and be motivated? That's possible but extremely speculative, as if he doesn't have motivation to win as much as possible already, keeping the 2900 goal in mind.

Or maybe he would be more stressed out having to play higher rated players. It could go either way, and in the end there is unlikely to be any difference.

I wonder if he could crack 3000 sparkles on his tutu in ballet?

In 2-4 decades, a 2850 will be about the same as a 2750 player now, due to ratings inflation.  Ratings are completely relative, and don't reflect actual playing strength at all.  

If you replaced every single rated chessplayer in the world with a 5 year old tomorrow, a handful of those 5 year olds would still maintain astronomical ratings relative to all the other 5 year olds, in spite of the fact that any average club player could come in and mop the floor with them all. 

I think that we should all just agree that Carlsen is perfectly suited for a sparkly ballerina outfit

btickler wrote:

Elubas wrote:

All that would do is exchange winning more often with beating better players. Sure, Magnus would earn more points beating 2850s. But he wouldn't win as often and would lose more :)

Maybe for some reason he would enjoy playing better players and be motivated? That's possible but extremely speculative, as if he doesn't have motivation to win as much as possible already, keeping the 2900 goal in mind.

Or maybe he would be more stressed out having to play higher rated players. It could go either way, and in the end there is unlikely to be any difference.

In 2-4 decades, a 2850 will be about the same as a 2750 player now, due to ratings inflation.  Ratings are completely relative, and don't reflect actual playing strength at all.  

If you replaced every single rated chessplayer in the world with a 5 year old tomorrow, a handful of those 5 year olds would still maintain astronomical ratings relative to all the other 5 year olds, in spite of the fact that any average club player could come in and mop the floor with them all. 

You're awfully noisy for a dead man, Beren.  Also, Beren Camlost would have more dignity than to reference sparkly tutus... ;)

No, he has too much to be positive about Carlsen.

Yeah so this totally depends on one's views on rating inflation. I happen to think it's false, and merely a tempting thing people like to say but without much substance behind it, but there are other places to debate that.

Not saying you yourself haven't looked at it carefully, but from what I understand it only seems plausible to believe in rating inflation based on superficial considerations.

Meanwhile, Grischuk has quietly passed Kramnik, Caruana, and Anand and sits at #3 in the world...

No worries though; at least we have cleared up on exactly where our disagreement lies. We can just leave it at that.

If the chess world were a closed ELO system, there would be no inflation.  It is not closed, though...every single day people pour into the system with their provisional ratings, and a subset of those lose a ton of games and leave the system never to return (far more than high rated players leaving the system).  So, when a player comes into the system and loses 100-200 ratings points and then bails...where do you think those points go?  They go to a handful of 1400-1600 players, who give them to a bunch of 1800 players, who lose to candidate masters, who get massacred by masters and grandmasters, who give them to Carlsen.

A rising tide lifts all boats, and Carlsen is treading water on top, but he can't walk on water.  There's nothing superficial about it ;).

What if the boat sank? Itg won't help much.

Ok: take this wonderful point you're making, and now apply it equally the other way.

What if for example an unrated player beats his opponent then no longer plays in tournaments? If we assume what you described "inflates" the system, then my scenario would deflate it.

I suppose you can speculate on why somehow unrated players who drop out would lose more often than win or something if you really want to maintain your position, but at least for me, given what I know, claiming inflation or deflation here is like being certain on whether a coin will land heads or tails --  such a claim would just be arbitrary.

Another way to handle the issue is to simply assess the quality of the moves played -- people like Dr. Ken Regan (also an IM) have done such studies using engines. It's not to say playing like engines is the only way to improve, but general accuracy of moves will correlate very strongly with being a strong player. Of course most people who do this kind of analysis have different methodologies, some better than others, but nonetheless I think this is the most reliable way to tackle the question of rating inflation. I'm not saying the issue is settled, but I'm not going to casually posit the existence of something without a decent basis behind it.

i.e. The creator of the ratings system clearly states that inflation (and deflation) are possible in his system.  It isn't just that the system is not closed and new players come in, lose points, then give up and leave those points behind to cause inflation...rating floors also cause inflation, as do one-off "adjustments" such as giving all women players (minus Polgar) 100 ratings points.

FIDE has no particular reason to curb ratings inflation unless it gets out of hand and begins to look absurd...after all, it's always a good news story when a new ratings highs are reached.  A minor amount of inflation is good for publicity, and inflation has been ~125-150 pts in ~40 years (estimated, but in line with studies done on various time periods)...pretty much perfect for continuously allowing new players to scale a slightly higher Everest, but not really enough to make the ratings appear nearly as relative as they actually are.

Elo states that inflation is possible, but hasn't happened. 

Nobody who has claimed inflation has happened has been able to state what that rate of inflation is. 

Arpad Elo, Kenneth Regan, and many other professional statisticians have argued that the rise in ratings is not due to a systemic problem, but rather the rise in ratings is due to increased ability! Elo states that in the 18 years since his system was used, he has not detected any such "drift" or "inflation". Regan has shown that if there has been any drift at all, it has been a very slight deflation of ratings! 

Again these are general comments, some of which I have already responded to. It's a little naive to simply assume our abstract dreams will actually happen in reality (things tend to be more complicated than we think). Again you raise interesting ideas in your first paragraph but they are general, the kinds of things that sound good yet in reality things might be more complicated or might not turn out quite as they seem from a distance. How could we even, empirically, confirm if and when these things happen? (Well engine analysis was my suggestion although you might not like that one)

Something a little more rigorous is needed to settle the issue. I don't think it's been settled either way necessarily but I am sceptical of inflation.