I finished up my report over the course of yesterday, and I found out a little more information about the Colts season, and why they were 7-9 with my predictor.
- Running the football contributed more to winning that passing the football did. The Colts, obviously, rely on Manning for a heavy dose of yardage. They were always on the top, or near the top level each week in passing stats. The Win % went down considerably when they didn't run the ball as well as before.
- The Colts never were above average in Time of Possession at any point in 2007. I'm not sure why that is. In 2006, it was painfully obvious, as teams ate up chunks of time running the football. Last year, though, that wasn't the case. The Colts only had 4 games with Time of Possession over 31 minutes. ToP was a key factor in the model.
The most surprising finding from last season was Week 3 vs. the Texans. The Texans had won 4 in a row (going back to 2006), and were above average in the four stats that had the biggest impact on winning in the model (Turnovers, Time of Possession, Rush Attempts, Rush Yards). The Colts were below average in all 4 of these. It was very lopsided. Obviously the model needs some tweaking. Here is a table with Win Probability for each of the Colts games last season:
Once the season got going, it predicted the Colts to win with much greater frequency. Week 17 was obviously assuming the Colts would play the first team throughout the entire game.
As always, any suggestions you may have that might help out, let me know. I'll be constantly updating throughout the summer, so it's in top form come September.