The Record so Far . . .
A couple days ago I glanced at the value for my average opposition rating for my correspondence games and wondered how my performance held up against specific rating ranges. I've still played few enough games on chess.com that I can quickly go through my spreadsheet record of my games and categorize the results.
As of October 2, 2010, my chess.com correspondence rating is 1698, and my average opposition is listed as 1644. Chess.com shows I've completed 85 games with 48 wins (56%), 32 losses (38%), and 5 draws (6%), but the number of wins and losses isn't really as cut and dried as that, due to unrated games as well as deciding where the cut-off is for a timed-out game to be considered aborted rather than a win.
My own kept stats credit me with 42 wins (49%), 38 losses (45%), and 5 draws (6%), and those are the games I'm using for this exercise. I also have another 11 games that were technically "wins" but which I record as aborted rather than including them in my results stats. These games include situations such as tournament or match opponents who have withdrawn from the competition after our game was scheduled or started, when a win has to be credited for purposes of the event but there is no effect on rating. None of those games are included as wins here.
In the table below, the percentage results within each 100 point rating range are forced to 100% for that range.
The bottom line? It's pretty much as you would expect. My rating is essentially 1700, and as can be seen on my rating graph (below) I haven't strayed too far to either side of that value for the past 50 games or so. In the above stacked histogram (forced to 100%) it can be seen that I've won a strong majority of my games within each 100 point rating range below 1700, and lost a strong majority of my games within each 100 point range above 1700. The red starts dominating for games with opponents rated above 1700, and I haven't even scored any draws against anyone rated over 2000. It's clear at which point I start having trouble. Now I just need to determine what the specific trouble is.
I don't know how chess.com determines the value for my average opposition. I haven't checked to see whether this value floats with real-time changes in the ratings of my opponents, or whether it's fixed. If it is fixed, I don't know whether they use the opponent's rating at the beginning or end of our game, or some other value. There can be a big difference, in either direction, in someone's rating when a game ends compared to their starting value.
Almost all of my games have been in either tournaments or team matches, which means I've played almost all my opponents twice-- once with White and once with Black. For each game I record both the starting and ending rating for my opponent, and for purposes of assigning a rating to the opponent I use whichever of the two values is higher. If I've played more than one game against the opponent I apply the higher rating for them to both games. They are usually either simultaneous games (same start date but probably different end dates) or immediately back-to-back games.
My rationale for using the higher rating is that for an established player the high value represents his/her capability, whereas the low value can result from a series of timeouts or some other non-ability cause. I've had a couple instances where one player's high and low values while playing me were separated by several hundred points due to a major series of timeouts. Obviously it would not be representative of the player's ability for me to apply the low value to him.