I believe that the number of games required to "stabilize" the ELO formula is quite high, around 500 games. (On FIBS, the formula stabilizes at 400 experience points, I believe. Experience points are the total of all your match lengths, so a 7 point match counts as 7 experience points regardless of the final match score.) Until you reach that number of games (or matches, here), your ELO adjustments will be increased by a factor that gradually declines to zero (or 1, I suppose, since it's a multiplier) as you gain experience. The great majority of players don't have that much experience. It's difficult to speculate about the effect this has on average ratings, but it must have some effect. I think the most important consideration is that successful players tend to stay and play more often than less successful players. (This is a correlation, not necessarily cause and effect.) Thus, a player who enters the pool and initially loses a lot of points may go away, or even just change his identity, leaving the low rating in the pool of averages. I believe this tends to drag down the overall average. This is one reason I based my earlier speculations on the average of the top 50 established BKR. The drop-out element is unlikely to be represented in that group! (Putting aside those few players who enjoy initial success, fly to the top of the ratings, and stop playing to guard that inflated status).