In Chess.com there is a bell curve of ratings at http://www.chess.com/echess/players.html. It shows an average of 1384 with almost everyone between 700 and 2100. I imagine that if one took all of the USCF rated players or all the FIDE rated players you would find a similar bell curve, probably with a different average and different extents. Assuming that all of the ratings are roughly bell-shaped then someone could create a mathematical formula for translating one rating into another. For any individual player the formula might produce very incorrect data but when averaged over thousands of players it should be pretty accurate.

So does anyone know if there are published bell curves for USCF/FIDE? Has anyone tried to create a mathematical translation? For those Chess.com players who do have official ratings elsewhere does there seem to be a reasonable correlation between ratings here and ratings elsewhere?

I think USCF has more 0-1000 rated players then Chess.com, making the bell curve wider, or back more.

Ratings are not absolute, they are relative to other people in the same rating pool. As such, it does not make sense to compare ratings from different rating pools. That's even before you get into the issue of correspondence vs. otb play.

Its an interesting question... but the ratings pools may not be equivalent (since otb club players who have more expensive uscf ratings tend to be stronger than casual internet players).  The best way would be to collect a large sample of ratings of people who have both uscf and chess.com ratings.

I'm about 1500 on chess.com and about 1200 in USFC

So there you go.

I guess it also depends on what times we are talking here: i doubt most players at chess.com ever play 2 hour games.Most people play 5 or even 1 minute games.

If USCF has tournaments for 5 minutes only or 1 minute,,then those should be the ratings  to compare.

I'm 959 USCF and 1840 Chess.com, but I don't think that I'm a very representative sample

There is no way to translate them, period.  They are based on completely different systems, and populated by completely different pools of players. 

It's about like trying to compare two glasses of juice, sight unseen.  You don't know if either is apple, or orange, or grape, you don't know if they are natural or sweetened, you don't know if they are fresh squeezed or from concentrate, and you don't know the exact size of either glass.

So how do you compare them?

Thanks for that link of the bell curve and other statistics. Comparing my stats for blitz and standard made me feel pretty good.