We’ve extended our sim to include the conference championships, committee selections, and playoff outcomes. This is a blend of two kinds of uncertainty – our usual power-rank based methodology and speculation about playoff committee politics. We model both and try to be transparent about which inputs are objective and which are subjective.
In this summary table the “Conference” column gives the probability a team wins its conference, based purely on M-P numbers. “Playoff” is the probability of making the playoff, which is speculation about the process conditional on play through the conference championships. Then “Title” is the championship probability, based (again) on M-P numbers, conditional on making the playoff.
We model the committee process by placing teams in tiers, based on what they accomplish between now and the end of conference championship play. This includes all teams we believe the committee will consider. Some teams show up in multiple tiers, representing different combinations of Ws and Ls in their remaining games.
The tiers reflect what we believe the committee’s priorities are. We are explicitly saying they will choose a team from a higher tier before they will choose a team in a lower tier. On the other hand, we do not pretend to know how they will select teams within a tier. To reflect this we simply randomize within the tier.
So the process and structure involves four steps:
- Simulate all remaining games
- Populate the tiers based on what happens in the simulation. These are the eligible teams.
- Start filling the bracket with the top tier. If spots remain, move to the next tier, and so on.
- Within a tier, randomly choose among the teams.
That’s it. Pretty simple, though obviously still an inexact process. And there is a lot we don’t know, especially related to how the committee will value a two-loss conference champion relative to a better one-loss team that does not win their conference. And what about the Big 12, which doesn’t have a championship game? All questions we wrestled with, and none with definitive answers.
A few more details…
Our first “tier” is an undefeated Alabama. But we also think Alabama is in the first tier if they lose to Auburn, yet still win the SEC Championship. Behind Alabama, in Tier 2, are Clemson, Ohio State, and Michigan, if they win out and win their conference championship. Tier 3 is a Washington team that wins out to win the Pac-12, one-loss Ohio State (that beats Michigan but loses the Big Ten East head-to-head tiebreaker with Penn State), a two-loss Big Ten champ, and a two-loss Florida that wins the SEC championship. Tier 4 is a one-loss Louisville and a one-loss Alabama team that loses the SEC championship (to either Florida or Tennessee). After some debate, we also decided to include the Big 12 champion, provided it’s one-loss West Virginia, or any other two-loss team. In Tier 5 is a two-loss Pac-12 champ and a three-loss SEC champ. Tier 6 is USC as a three-loss Pac-12 champ, a three-loss Big Ten champ and a two-loss West Virginia team that wins the Big 12. Finally, our last tier is a three-loss ACC champ, a three-loss Pac-12 champ that is not USC, and an undefeated Western Michigan.
Within each tier, our ordering of teams is random, which helps us bake-in some much-needed uncertainty to the process. We may be overstating how much emphasis the committee places on winning a conference, but at this point we really don’t know. Last year, all four playoff teams won their conferences, but things were (at least relatively) straightforward; it’s hard to know how the committee will decide if there is a lot of carnage the rest of the way.
Happy to hear about disagreements, and may tweak the subjective bits as we learn more about the committee. It’s also possible we a contingency here or there – let us know if you see a hole.