Scraping Canabalt scores off Twitter, ramming them into Mongo and processing in R. The results are not vastly revelatory, but it's a nice account of the process of storing, processing, and representing big data.

"...in honor of Canabalt's first birthday (back on September 1st), why not write a guided tour of how everything works instead?" Lots of good tidbits on here; the stuff on hitboxes is particularly well observed and explained.