Fencer,
I would have to agree with the comment that drawn matches do not count. I have had quite a few drawn games. In many cases the difference in ratings was quite a lot (e.g 2000 vs 1300), in which case I would expect to increase the lower rating and decrease the higher rating. NEVER has this occurred. I suspect that there is a defect in the code somewhere causing this.

In addition I would suggest that winning a series 3-2 (which should score 0.6-0.4 where a draw is 0.5-0.5) should score less than winning 5-0. I haven't enough data to tell whether this is the case, but can you confirm that it should be? I suspect that the same piece of code would be used in this case and therefore it may not work either.

If necessary I volunteer to play someone in a game of something where our ratings are dissimilar and engineer a draw to test this hypothesis. Just invite me to a game and alert fencer so he can monitor the ratings.

Just trying to make sure it works!