With Eric Chemi of CNBC, I calculated an ensemble prediction for the Super Bowl that combines my numbers with other trusted sources. This ensemble predicts a slim 0.46 point edge for Seattle, which gives New England a 49% chance to win.
Here are the sources I used in the ensemble.
- The Power Rank – I used my rankings that combine calculations on margin of victory and yards per play.
- Advanced Football Analytics – Brian Burke uses yards per play and offensive turnover rates in ranking teams.
- Numberfire – Rankings based on an expected points analysis of every play.
- Football Outsiders – Aaron Schatz uses the idea of success rate on every play in his DVOA rankings.
- Inpredictable – Mike Beuoy takes market data and uses regression to rank teams.
- Massey Peabody – Cade Massey and Rufus Peabody use football play by play data and weight recent games more.
- Prediction Machine – Paul Bessire simulates the game 50,000 to come up with a prediction.
- Sagarin Pure Points – Jeff Sagarin, who developed his ranking algorithm back in the 80’s, uses margin of victory in games. Not modern but still useful.
- Line – the point spread from the markets.
I wrote about expected points and success rate used in DVOA in this article.
You say Sagarin in not modern but still useful. I will argue that it is the best of the bunch you decided to use.
My research for thepredictiontracker.com has showed that all those systems using play by play data do not add enough to a simple systems using just scores.
And as you can see from http://www.thepredictiontracker, an aggregate of about 75 systems has New England a 0.51 point favorite.
Thanks for the insight. Would love to hear more about that research.