Out of sheer curiousity, I have questions for chess.com's staff about how your rating adjustment algorithm works.

When playing multiple games, I often consider what is at stake from a ratings standpoint before I agree to a game (e.g., "game x").  Since I play multiple games, I understand my rating goes up or down depending upon whether I win or lose other games. When my rating does change for this reason,  I've noticed what was at stake for "game x" when I agreed to this game has also changed.  If my rating has improved, what's at stake for "game x" changes as follows:  Winnings are diminished, losses are increased.  The inverse also appears true if I have lost rating points.  Why does this sliding scale exist?  Why aren't you using an algorithm that protects the initial stakes for "game x" instead?  Is the way this works similar to USCF and FIDE rating adjustment algorithms?

http://support.chess.com/Knowledgebase/Article/View/16/0/how-do-ratings-work

does that help?

Your rating is an estimate of your playing ability. It is calculated based on your results. When you win or lose a game, the system acquires new information and updates the estimate on your playing ability accordingly. Presumably, your rating becomes a more accurate estimate each time new information is acquired.

When making this update the system uses the most accurate information available to it. Since your current rating (and your opponent's current rating) are the most accurate these are used rather than your ratings at the beginning of the game.

The amount your rating will go up or down as a result of a win or loss is not a wager that you make at the beginning of the game, it is a mathematical estimate of your playing ability.

Erik,
 
Thanks for your explanation.  Your site uses the Glicko System, which effectively layers a 'ratings deviation' (95% CI) over the Elo system the USCF and FIDE use.  Straightforward enough, and glad to know you were involved implementing it for chess.com.  Simply explained, the better chess.com knows you (e.g. the more you play chess on chess.com) the better SWAG they can make about your true rating.

Loomis,

I'm glad for your comment "The amount your rating will go up or down as a result of a win or loss is not a wager that you make at the beginning of the game...".  That is still exactly how I view it regardless.  Aren't ratings (as imperfect or precise as they are) exactly what each one of us wagers every time we play a rated OTB or online game?

Fezzik,

I'm also glad for your comments.  I'm happy to know chess.com is consistent with USCF and FIDE in using post-event rating values to calculate rating adjustments.  Still, they do differ in their ratings-adjustment calculation.  USCF and FIDE use Elo to calculate rating adjustments.  Chess.com uses Glicko.  I'm curious to know what inputs each uses for their post-event ratings-adjustment calculation.

Again, thanks to all of you for your posts.  I do believe I have a much better idea what the answers are to my original questions.  When I see "Show Ratings Adjustment" I now take it to mean "show me an approximation of what my rating adjustment will be for this game (using current data)".

This is Mark Glicksman, I think he is a genius.

I always wondered how the rating system was derived.