chess.com ratings are deflated against USCF

Actually my chess.com ratings are very old though, so who knows what I'm rated right now.

I'll try to get a current rating on here in the next few week or so and see what happens.

The naive OP still didn't understand that online chess is not reliable. A friend of mine 1800 USCF lost here against another player 1300, for 6 games in a row. I hope we all understand what it means, and it is not rating deflation.

And I'd like to add that you'd find a lot of perfect +1900 player to play against, but also in that case they are not human or they are half-human and half.... which would still make the rating comparison (without a real huge random sample) totally ridicolous.

You might say that online chess at chess.com is not reliable but you can also say online chess is reliable at some other sites which do not use this nutty rating system or I should say this aberration of a good rating system.

What's wrong with this rating system?  Don't want to search 5 pages of posts for what you've said in the past.  From my skimming it has to do with starting a player with a certain rating?  But RD allows it to change so fast in the beginning.

Well, OK, technically you can refuse to rate a game between two unrateds ... although it's kinda silly to play rated games and still be unrated.

I have a rating.  If I play an unrated, my rating is obviously going to change.  But how do you determine the assumed rating of my opponent, to determine the change in my rating?  I can think of 3 ways to do this:

1) Assume the new player has a default rating, such as 1200.  (With the initial RD of 350, you can say there's a 95% chance that this rating is within 700 points of the player's true rating.)

2) Give the new player a rating based on his opponents.  (Since I'm his only opponent so far, his rating would then just be some formula based on my rating and the outcome of the game.  If someone's first game is against a player rated 400, their rating will be 800 or something.)

3) Wait until the player has several games completed and then calculate everything retroactively.  (I don't think the chess.com servers would appreciate this.  And there are players who play one game and leave the site.  You can't assume multiple games will EVER be played by all players.)

Perhaps it is to do with why you have come here and what you enjoy. Someone who wants to find other competant chess players, and to play, possibly doesn't care a whole lot about their rating. If they play a dozen or two dozen games before establishing a chess.com rating which reflects their strength compared to the rest of the chess.com community of players I doubt they will care.

If, on the other hand you are someone whose pride is tied up with rating - maybe because you have invested time and effort into your achievements in the world of chess or maybe for some reason more related to your psyche - then you won't like the idea of accepting a low rating or playing for a time at a level which does not suit your sense of your standing in the chess world.

The thing I particular like about this site (apart from it being very well run I mean) is its inclusive nature. It brings high class players into close contact with patsers and outright beginners.  One benefit of the rating system run here is that a newcomer can make a start and see how they get on; can be motivated by the desire to improve their rating; and can readily see the results of their play from the very first game.

Perhaps that would happen with some sort of "unrated" start but perhaps not.  People starting out with a background of chess played at home or at school or in a club will get a lot of encouragement from friends and family.  Someone whose main chess experience is this site needs as much encouragement as the site can offer.

I would, I think, prefer a rating system which works for newbies than one crafted for 2200 plus players. 

why would anyone want to "estimate" what their blitz rating is when they can just play and find out?  The fact is if you use a population of people who have blitz ratings higher than their uscf ratings by 200+ you are going to get an opposite conclusion. And don't tell me they don't exist I see a lot of  low 2000s uscf players who are 2100+ blitz.  Anyways I think you just trying to justify why you have a low blitz rating compared to your USCF just to boost your self esteem.  A 1550 blitz on this site definitely does not play like an 1800.  

Both Glicko and Elo have their advantages and disadvantages.  Elo gives more weight to your most recent games and is more volatile.  Note that volatility can be controlled by the constant used, and the constant used by Chess.com varies with number of games. 

Do you people lack critical thinking skills? I'm obviously aware of what online chess is. Nevertheless, there is an average ratings difference between the two pools. That's just a mathematical fact. Why pretend otherwise?

Why do you lack reading comprehension skills? I have more than 'ten' samples. Why don't you read what I write, before you start trying to argue? I wish there was a way to filter out stupid posts from people who didn't even bother to read the thread.

Adam swings out at anyone who challenges his position. Instead of attacking everyone else, perhaps he could read the comments, especially the ones pertaining to statistical methods, and perhaps learn from them?

Lol, compare the numbers for yourself. You'll see that I'm right.

I personally wouldn't mind actual numbers here. As in have a column of USCF ratings and corresponding chess.com ratings. For one thing, this would give a better idea of the spread that is covered here. A 600 USCF is unlikely to have a blitz as low as 350 if for no other reason than that there are no players that low rated on chess.com. As well, and more importantly, we would actually know the sample size instead of "I have more than ten samples, {insert condescending follow-up here}."

Also, you know that USCF tracks "Quick Rating," right? Wouldn't that make more sense for your sample? I understand you have claimed that it is simple to transition from longer time controls to shorter, but regardless of the accuracy of that claim, it just seems simpler to cut out the middle-man.

Anyway, that's my (likely inane) two cents.

Yeah, there should definitely be a correlation.  I don't think you've seen enough data to claim anything definite yet though.

There very probably is a correlation however... 

@ Motty
Well, I can almost guarantee a correlation because stronger players have higher ratings in both populations... so... yeah.

A nice and tidy linear correlation across the board... +250, is of course too simple.

I will guess the +250 is within 100 points of the exact correlation?

I suppose I have a rating drop to look forward to then when I start in on live games heh.

I apologise for mis-remembering your original post.  Still, it is not a random sample because some (I won't assume all) are in your personal circle of friends/acquaintances/opponents/etc.  We may just have to agree to disagree on this.

I do a find it a little ironic that you take me to task for making unwarranted assumptions in one paragraph, and then turn around in the next one and assume without any apparent basis that the other folks reporting different anecdotal results simply did it wrong.