This is a python module to scrape basketball-reference.com and convert various stats into usable data structures for analysis.
Here is a link to a sample IPython Notebook file demonstrating the library.
Beautiful Soup (version 4 or above)
Pandas (0.11 or above)
Still developing the API. Right now you can get a list of all player overview urls, generate a list of game log urls for a given player, and convert that list into pandas dataframe.
players.json was generated on 03/09/2013 by buildPlayerDictionary() and savePlayerDictionary(). It is a good way to jumpstart your analysis and can be loaded with ... loadPlayerDictoinary(). Note that it's a pretty large (13M) file. I'd recommend building your own, fresh copy. Note that it takes about 10 minutes due to spacing out the web requests.
I'm considering turning this into a class, instead of using a dictionary, so one doesn't have to pass around this dictionary all the time. Hesitant.
Local Database construction.
Extract other key information from the player overview page- position might be an especially useful variable to use for supervised learning and height/weight is a clearly important variable as well.