Hi,
Please have a look there, message #3
http://www.massiveassault.com/forum/viewtopic.php?t=4211
Not all the points deal with scoring (but also with trial players and players with less than 15 games, whose score is unreliable).
Glicko is not as mysterious as it... sounds?!
About the point from Storm440: too much movement. Well, just read quickly the white paper from Glickman (*), and you'll see that a period should exist for calculations, and NOT calculating the score at each game. This, undoubtly, will smooth the score change. Silly example: 4 games {W,L,W,W}; instead of having for score {up, down, up, up} you'll have a smooth one {up}. The more game you play within a period, the more accurate your score would be: you can guess it by yourself, although it is also written in Glicko's explanations.
Well, if you still need this "real-time score" with its up&down, here's a suggestion: just add a "average score" which would be the average of all your previous "real-time scores" from the whole previous week/month. This new "average score" should be more accurate, isn't it? Anyway, I personally would find this solution weird since Glicko IS smoothing changes IF you take a period.
Olivier
(* A link could be find there:
http://www.massiveassault.com/clans/nwo/allratings2.php where you'll find another way to smooth drastically the score change: individual score by planet (score which moves a lot, as here in MAN2(?)), and then an average score for whole planets (score moves less, since you divide the theorical Glicko's delta/move by the numbers of planets).