The NFL Game Predictor is a program I wrote to predict the scores of NFL games based on stats from previous games. I wrote the program in Java and it parses game stats from the web and uses them to build a model to predict NFL games. In addition to analyzing stats from games, the predictor also takes things like bye weeks and upcoming opponents into account. Originally the predictor used a simple linear regression model, but since then I have improved it by evolving a neural network to predict the scores of the games based on the stats.
The program has been able to pick games correctly straight up 65-70% of the time while picking games against the spread correctly just over 50% of the time.
Here you can view predictions for every year since I started the program:
My Patriots Blog