Well, it looks like shit at the moment, but it's almost done. All I need to do is set up automatic fuckingoff of ridiculous outliers (search the market for aqua blood for an example), a cron job to generate the right graph files and another to generate the actual graphs (as well as some tweaking to the actual graph generation process).

The finished graphs will look something like this (except with labeled axes and, in the case of multiple sets of data, a key):