I think that you are correct, the last rating of each player is used to make a final determination. This makes sense with new players joining all the time and games often taking months. Imagine you started a game with a 1200 player, six weeks later he's rated 2200 and you only get a couple of points for beating a beginner. Not nice.
Further reading:
http://www.chess.com/article/view/chess-ratings---how-they-work
Ratings are an estimate of playing ability. The estimate is never perfect, the rating systems just tries to make it more accurate every time new information is available.
The completion of a game is new information. The information is whether you won or lost and the estimated playing strength (rating) of your opponent. When considering this new information, it makes the most sense to have the most accurate information available.
Since the rating system makes ratings more accurate with every completed game, the current rating is always the most accurate estimate. For this reason, it makes sense to use the current ratings in order to determine the rating change. That way we make as accurate as possible updates to ratings.
PrawnEatsPrawn wrote: "Imagine you started a game with a 1200 player, six weeks later he's rated 2200 and you only get a couple of points for beating a beginner. Not nice." Very true! What prompted me to take notice was a game I was winning, and originally the rating change would have netted me 100+ points gain, however after the mid-point of the game I noticed his rating had slipped, and now I was going to earn less than 80 new points. Not nice either, but that's just spilled milk.
Loomis, thanks for the most clearly put explanation of why the current system makes more sense. You write clearly & I appreciate that!
It appears to me that in the turn-based ratings the change is based on your and your opponent's ratings at the time the game is completed. Is this correct? If so, why? Shouldn't the ratings used to calculate the change be those of the players when the game was begun? I know I shouldn't care, but I can't help wondering...
And yes, Erik, I did read this. One of the steps given says "Step 1. Determine a rating and RD for each player at the onset of the rating period." Which seems to suggest using the ratings that were in effect at the beginning of the game, not when the game finished. Just curious.