New Year, New Model

I spent the past week refining the modeling algorithm for the upcoming season.

There are three main changes:

• I now predict two outcomes: Probability that the favorite wins the game (1) and the probability that the favorite beats the spread (2)
• The model now uses historical match-up information (for example, what is the likelihood that the Steelers will beat the Eagles given that the Steelers have won the last 4 match-ups?)
• The model adjusts for "1/2 point" increments in the vegas spread. Those half point situations have a MAJOR impact on the model, success rate is now up to 65%.

Mike