Week 6 Recap and Week 7 Picks

Well, despite probably hundreds of hours of development time, Weeks 4-6 of model predictions had a simple error that based the predictions on 50% of last year's data and 50% this year's data. I caught the mistake when I noticed that Baylor was predicted to lose to Kansas State by a wide margin; not that this isn't possible ... it just didn't pass the common sense test. When I identified the error and reran the model, the results were more in line with what one would expect.

This error tended to minimize the best teams' expected wins and increase the worst teams expected wins. For teams like Louisville, the difference in expected win counts from Weeks 6 to 7 are significant.  For other teams, it isn't as noticeable.

Click on a team to get a close up graph of that team's expected wins over time in the lower chart, or use the filters between the charts to sort or filter on conference or team.