After an eventful “offseason” (I’m not sure either Cade or I understand the concept of downtime), football is finally back! In the last six months, Cade got himself hitched to a wonderful woman, while I took advantage of my lack of commitment and backpacked for 10 weeks in southeast Asia, all while keeping up with my work. This will be our fifth season working together on Massey-Peabody, and our picks have turned a profit every year, a streak we hope to continue this year.
The last few months, I’ve done a lot of work trying to improve the model. We love the simplicity of our model–we only use a few key statistics and do an excellent job contextualizing them–but I’d been chomping at the bit to dive deeper. The biggest change I looked at was breaking down our “scoring efficiency” statistic–a catch-all for everything not measured in play-level performance–into its component parts. What drives scoring efficiency? Red zone performance, special teams, turnovers, penalties, in-game coaching decisions. We know that the predictive value of each of these components doesn’t correspond exactly to their relative weight in scoring efficiency (just think of a 95-yard fumble return for a TD). Two teams can have identical scoring efficiency numbers that are driven by different facets of the game. My idea was to predict each component separately using a framework of expected points gained/lost, then aggregate them into one number measuring the points we’d expect each team to gain or lose based on these auxiliary components. I succeeded, but found that (a) scoring efficiency still has value, and (b) the gains were trivial, and not enough to justify the degrees of freedom. So the models are still very similar to last year, with some minor changes I won’t divulge. 😉
One change I will share is that we are now using pace of play in converting our college power ratings into point spreads. Our power ratings now represent the expected margin of victory between a team and an average team assuming an average number of plays run.
This season, we’ll be providing the following free content:
- Weekly power ratings and data visualizations for both college and pro football.
- Weekly picks against the spread for both college and pro football.
- Weekly NFL season win predictions and playoff/division/championship probabilities based on our simulation engine.
- Weekly content/commentary in the Wall Street Journal online edition and semi-regular content in the print edition.
- Weekly (awkward) videos on The Linemakers where I discuss the week’s games with Kenny White.
If you have any comments/questions/requests, don’t hesitate to contact us by selecting the “Contact Us” option under the “About Us” heading.
Thanks for reading, and good luck this season!