The Web-based Flash game Canabalt, whose scores have been analyzed by R before, is now available as an iOS App. Because the app is configured to work on three different platforms: the iPad, iPhone and iPod Touch; and because players are invited to tweet their best scores at the end of the game, like this:

the Twitter stream again becomes a great data source to analyze the distribution of Canabalt high scores, and now also to compare performance on the various platforms, and even to analyze individual death types. 

R user Neil Kodner has done just that, by writing Python code to scrape data from such tweets stored in a MongoDB instance, and then using the plyr package in R to aggregate the data to create some interesting visualizations of the scores, such as this one by platform:

Surprisingly the iPhone, despite its smaller screen, is the platform upon which the higher scores tend to be reported. (It's also the most-played platform, which surely skews the results somewhat — experts on extreme value theory are welcome to chime in!) The deeper analysis by type of death is also interesting: my own demise in my high-score game “somehow hitting the edge of a crane” is one of the rarer ones. Check out the full analysis and code in Neil's blog post linked below.

neilkodner.com: Visualizations of Canabalt scores scraped from Twitter