Some days ago, my office-mate YFG sent me a link which was about a competition for predicting results in college football games. The competition is set up by TopCoder website and main sponsor is ESPN. In this competition you should predict this year’s results (2008 season) based on statistical information of last three years (2004-2007). Although we do not know anything about (American) football, I think we can still use some machine learning methods to come up with quite good results.
In my opinion such challenges are a good way for ML guys to test their methods and learn some subtle points which are really important in any practical application. Moreover it will help you to be more enthusiastic and to feel you are working on some cool stuffs. Besides those “academic” benefits, we can also earn money up to 70,000 USD. So lets ROCK it! Here is a diagram which shows deadlines for each sub-season predictions
