The Massey-Peabody bracket reconciles politics and analytics, showing that predictive analytics can and should be used even in a political process. We begin with the traditional Massey-Peabody Power Rankings, a predictive analytics model built to identify the “best” teams. We then impose a few political constraints in order to also consider which teams are “most deserving”: 1) Every team starts the season equal (hence we use Massey-Peabody’s “hybrid” rankings, MPh, detailed below), 2) No more than two teams per conference (and at least three conferences represented), and 3) Priority given to conference champions and head-to-head winners. We do not consider win-loss records, as they are notoriously unreliable indicators of team quality.
Massey-Peabody bracket if played Nov 8
1. Alabama, 7-1, SEC, MPh #1 (MP #1)
2. Ohio State, 7-1, Big 10, MPh #4 (MP #5)
3. Auburn, 7-1, SEC, MPh #5 (MP #2)
4. Oregon, 8-1, PAC-12, MPh #6 (MP #3)
This is getting tougher. And more subjective. How closely ranked must two teams be in order to consider their head-to-head performance (the committee uses a 6-team band when doing their final rankings)? When head-to-head is considered and matters, does the loser fall, winner rise, or both? Turns out politics are more complicated than “simple” analytics. But good numbers still help us do better.
Our top seed is easy, with Alabama being a clear #1 whether or not you consider pre-season expectations. Michigan State is #2 in the Hybrid rankings but lost to #6 Oregon, so gets bumped out. Similarly, newcomer Oklahoma is #3 but lost to #9 TCU so also gets bumped out. Ohio State slides into the #2 spot, but has to face Michigan State in East Lansing this weekend. Auburn picks up the second SEC spot, landing at #3 after their brutal win in Oxford. Oregon grabs the final seed on the heels of a big win against their nemesis, Stanford. Mississippi State (MPh #7, MP #9) and Florida State (MPh #19, MP #4) remain on the outside looking in. We are unfazed by their undefeated records and – in FSU’s case – lofty pre-season expectations, instead caring only about what their in-season performance suggests they will do in future games.
Background: To be more politically palatable, we created a modified version of our rankings that does not use priors. In these rankings all teams start the season on equal footing and are never affected by our pre-season expectations for them. (Tip: This is a bad way to bet.) But we do use priors for strength-of-schedule considerations. Hence, while Marshall and Alabama begin the season equal, we know Alabama plays the tougher schedule. It’s a fudge. But it allows us to maintain some realism while staying “pure” in our evaluation of this season’s play. Below is our current “Hybrid” Top 15, along with the rankings from our full model (MP).
|Rank (Hybrid)||Team||Hybrid MP||Rank (Full)||Full MP|