We figured we should offer a bracket for this year’s college playoff. Since the playoff committee is announcing their first rankings next week. And, you know, since nobody else seems to have an opinion on the matter. The challenge is that we want a bracket informed by analytics – predictive analytics – but one that is politically viable. After all, this bracket is not about simply seeding the “best” teams (those you’d take in the sportsbook), but some balance between the best and the “most deserving”. That means we need to consider things like conference championships, head-to-head performance, geographic diversity, etc. Yet we don’t want all of that to drown out the analytics. Tricky balance, but a good challenge. We’ve come up with the following, which we’ll update weekly going forward.
Massey-Peabody CFB Playoff Bracket (if played Oct 25)
Background: The Massey-Peabody bracket is a collaboration between predictive analytics and political reality. We begin with our traditional Massey-Peabody Power Rankings and then impose a few constraints: 1) No influence from pre-season expectations (hence, we use Massey-Peabody’s new “Hybrid” Rankings, “MPh” – see discussion below for more detail), 2) No more than two teams per conference, 3) If more than 2 teams from a single conference are in the top 4, we give priority to conference champions, then head-to-head winners, and then the higher power ranking.
1. Ole Miss, 7-0, SEC, MPh #3 (MP #7)
2. Alabama, 6-1, SEC, MPh #1 (MP #1)
3. Ohio State, 5-1, Big 10, MPh #4 (MP #2)
4. TCU, 5-1, Big 12, MPh #6 (MP #10)
Commentary: Ole Miss is no surprise – widely seeded this high, they show up #3 in our Hybrid rankings but get the top spot because of their head-to-head win over Alabama. Alabama gets the 2nd and final SEC spot because of its #1 power ranking (regardless of whether we consider pre-season expectations). Ohio State is the top-ranked non-SEC team, in the top 4 on their own merits (and even higher if we were allowed to consider preseason expectations – they are widely under-rated at the moment). TCU pulls down the final spot, with its higher power ranking nosing out deserving 1-loss teams Michigan State (MPH#7) and Oregon (MPH#9). Georgia (MPH#2) and Mississippi State (MPH#5), are the biggest victims of the system’s (likely) political/geographic constraints.
More detail: To be more politically palatable, we have created a modified version of our rankings that does not use priors – all teams start the season on equal footing and are not affected, up or down, by our expectations for them. (This is a bad way to bet, btw – some of our edge in CFB comes from proper use of priors – but it seems reasonable to set them aside for seeding the bracket). One problem with setting aside priors is that we lose understanding about a team’s strength of schedule. So for example, if we run an in-season only version of our rankings, Marshall comes out #1. To add some realism (after all, the committee is trying to consider strength of schedule), we created a “Hybrid” model (MPh). This model sets aside priors for the team, but incorporates priors on their opponents. That means Marshall opens the season on the same footing as Alabama, but also that we know Alabama plays a 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. This is what our “Hybrid” Top 15 looks like right now, along with the rankings from our full model. We are in the process of posting a full set of these rankings, along with two additional versions, on the site.