What is the average rating in the USCF?

USCF doesn't appear to be using a system that is based on a straight bell-curve, but one that is weighted in some way.  So a 1200 player in USCF terms would not be comparable to a 1200 here on Chess.com.  If you were to graph the two systems, they would look very much like this (because these are snapshots of actual data):

(the orange graph is USCF, and the green one is Chess.com, naturally)

As was pointed out before, apparently the USCF includes scholastic players in the graph, which could partly account for the "double-maximum" geometry of their curve.

If I recall correctly, the USCF also gives new players a different provisional rating, too (Chess.com's 1200).  Not entirely sure, but the graph seems to indicate that the second maximum is somewhere around 1600, which would indicate something around a 1500 provisional rating for starting USCF adult players.  Can anyone verify this?  They also seem to have artificial brackets based around a modulus-100 (i.e. people cluster around each "100" points of rating, because movement into the next bracket can't happen until the next tournament).  I believe someone has already pointed this out.

The provisional rating is given after the first tournament, not before. If all games are lost then the rating is 400 points below the lowest-rated opponent; if all won, then 400 above the highest-rated opponent. Otherwise, the first rating is whatever the tournament performance was.

Afterwards, normal rating calculations apply. The rating remains provisional until 25 games are completed.

This produces much less inflated ratings than the chess.com, where each new player injects 1200 points into the pool.

In most tournaments I've seen around my neck of the woods, no strict enforcement of the lower limits of the brackets was done. People can play up when they wish (usually a small extra fee applies).

 That was a good guess, too.  Contrast that to Chess.com, where the 75th percentile is called 1360 or higher.  That seems to suggest that 1360 (Chess.com) = 1600 USCF.

 This "pool effect" is small - the actual number of points can go very high under Glicko (technically it could be infinite) because the curve doesn't end at the top, even if the bar is too small for a person to fit there.  The true cause of inflation is because people don't stop at where they start.  Most players will reach a higher rating after some time and never go much lower (they don't get worse at Chess - they just stop playing).  But anyone who gets a win against them continues to increase.  The average rating goes up as a result.

I'm not sure what it means that they can't move into the next bracket until the next tournament.  I thought this "clustering" was because of artificial rating floors.  A player given a rating floor of 1800 cannot fall below 1800 regardless of performance.

Okay, so I downloaded USCF cumulative ratings from 2009 (this includes all players who played at least one rated game in 2009). Counting only active, established ratings (25 games or more), I got this:

Average rating: 1183.  Median rating: 1163. 

Counting all active ratings, including provisional, I got:

Average rating: 855.  Median rating: 716.

I believe one shouldnt even have a published uscf rating if its under 1200..... having a fide rating used to have a lot more importance/prestige back when they didnt publish ratings under 2000 . At the very least maybe there should be separate "scholastic" ratings from adults/non-scholastic ?  I recall when the average uscf tournament rating was 1500-1600....... now the flood of scholastic players have driven it down. When did uscf stop using the elo system and start using glicko ?  I am not even aware they did.

I think this makes sense... no offense intended to anyone below 1200.  If you're a beginner, it seems like playing your friends or in unrated tourneys would be enough.  To have a national organization rate you, it makes sense that you'd want to be a more serious player.

It's just that chess skill is somewhat esoteric -- someone who has never seen a game may mistake even random moves as good play... compare to something like piano where hitting random keys and anyone could tell it's not a song.  So the USCF accepts their member dues and rates them. 

To be honest I don't know how anyone could be rated below 1000 without simply making completely random moves some of the time.  Of course we all started out that way, but tournament chess you'd think would be a little more serious...

I like the rating system as a gauge of ability. At the time of my first and only tournament, the elo system was in use. I didn't get a rating because I had poor results (come on! it was my first tournament); besides, the rating I would have recieved wouldn't have been an official rating. The rules of the ACF back then demanded a player participate in at two official tournaments before "locking" a player into the system and allocating the player's given elo rating. It seems this method of control managed the rating system better, in regard to having masses of < 1000 rated players. I think the average player (who knows the rules, basic tactics, and general motifs of a chess game should play at around the (elo) 900 - 1100; anything beneath this doesn't make a lot of sense when determining player strength.

Some info here http://www.glicko.net/ratings.html.

Edit 2010 07 28: Too quick on the draw.

It is Wikipedia, information can be suspect.  http://en.wikipedia.org/wiki/United_States_Chess_Federation 

"The current rating system as implemented by the USCF is still an Elo rating system, but with a sliding K-factor. There is an ongoing discussion within the USCF Ratings Committee of going to Glickman's Glicko-2 system in the future.".

"If I make $30,000 a year and Bill Gates makes in 1998 $45 billion, the average is almost $45 billion."

apparently i've been doing averages wrong all these years.

$45,000,000,000 + $30,000 =
$45,000,030,000 / 2 =
$22,500,015,000 (which is not almost $45 billion)

That is so not true it's hilarious.

Also, from personal experience, I'd say the average tournament is not actually a 1500-1600 as many people seem to suggest. In small tournaments, it's generally around a 1200-1300 rating average. Even in large tournaments, it's closer to a 1400-1500 average in most cases, to be honest. I guess it depends on where you live. Although, that brings up another question. Could certain regions be under-rated/over-rated? So that a 1500 in one place equals a 1800 in another?

Different regions in the US are slightly different I've heard... at least from a few people that have moved into my state.  Nothing like a 300 point difference like your example, more like 50-99 points.

I would guess a state with a big scholastic scene would deflate ratings while states with a small pool of players that don't play outside themselves may tend to have inflated ratings?  I'm not sure how you could test this.

Rating pools are funny like that.

I got my initial rating playing the ranks of extremely strong scholastic players at the Mechanics in SF and East Bay Chess Club. Every tournament I would end up playing Naroditsky, Shankland, Greg Young, etc... then I would go to something like the National Open in Vegas and find myself at the near the top of the crosstable in my class.

And again, if you separate kids from adults, the average scholastic rating would seem to be around 600, while the average adult seems to be around 14-1500. It's the overlap, and the fact that the kids outnumber the adults, which bring the average down. This is consistant with what I said earlier.

I'm just amazed by the people who keep saying that the average at the adult clubs they attend is 1200. Kids may outnumber the adults in USCF, but around here, only the better kids (1100+ ratings) show up to adult tournies and clubs, so the average ratings at adult tournies are usually around 1500. I still remember showing up to one of my first tournaments rated 1300 and getting clobbered, because my lowest rated opponent was in the 1600's. I think there were maybe 2 kids at that one, and they both rated higher than me.

 All you do is add all the numbers up, and then devide them.

Of all the online sites, Chess.com is the most accurate reflection of true USCF playing strength.  Mainly, because it's the most conservative.  My rating in World Chess Live is 1469; Chess.com is 1340.  I believe based on the games I've played over the board, that I'm playing at a 1340 level as opposed to a 1469 level.  Other sites like Pogo.com rate everyone at 1500 to start as opposed to 1200 here.  That adds 300 points to every player's rating.  I firmly believe that most entry level adults start out around a 1200 level.  Chess.com does a good job giving this estimate.  My Chessmaster GM Edition rating is about 1500. 

Chess.com, and all Glicko-type systems are strictly mathematical.  There are no floors, or stair-steps, and the curve represents a smooth distribution of all players.  In the older systems, ratings rely greater on performance (win-loss ratios, bracket mastering and progression, strength of opponent, etc.).  So you could typically gauge performance better under Elo/USCF (its plus side) because it identifies players as belonging to seperate peer groups, or sharing the same step on the staircase.  This is much harder under Glicko, because the RD must be specified as well to determine if a player's strength is accurate (I.E. Glicko has the geometry of a hill, while Elo is more like a staircase.  We know under Glicko, that someone is very near a certain point on the hill, but they may be a little higher or lower than that point.  Under Elo, we know they are somewhere on a certain step, but many other players are on that particular step as well).  Glicko's plus side is that a rating can be found quickly (after just one game, perhaps) and the rating becomes more accurate as more games are played.