IMPORTANT

I'm not actively developing this code any longer!

My current work on machine learning in march madness is being done in Scala in this newer repository.

ncaa

March Madness simulation and prediction object model and helper modules

Description

Supplied files facilitate the scripting of NCAA basketball simulation and prediction software. They reference a sqlite database on my server.

Files

Object Model

ncaalib.ncaa Object model written in Python using SQLAlchemy. Database stores information about games, players, and teams in a rich, interconnected fashion. Complete documentation can be found in this file's docstring. Include this module in any script that needs to access DB. Defines several convenience functions for searching the database as well. Can be run in interactive mode, in which case script will automatically create the variable session for quick CL investigation of the DB. See full documentation on my blog.
dbmgr.py Command line utility for managing the database. Provides some handy routines for mainting the database. Relies heavily on output from scrapers (also provided in this repo). Run python dbmgr --help for a complete list of options.
svmexample.py Demonstration of model training and Tournament simulation using data from the database.

Scrapers

Scrapers written for CasperJS. Not optimal for speed, but was at least convenient. Also easier to work around potential impediments in Casper than in Python.

scrapers/ncaaPlayerStatScraper.js Scrape stats.ncaa.org for basketball statistics by player, store in CSV, process with dbmgr.py.

License

Freely distributable under the MIT license.