Building a College Football Playoff Predictive Model
As it currently stands, there are five months and change until college football selection Sunday. For Buckeye fans, this day represents a 50% probability the afternoon will be filled with either joyous elations and high fives (given the pandemic, air fives will suffice) or tears of sorrow. The type of sorrow an average Buckeye fan can experience on a pain scale is between a Michigan loss (low probability) and the refs botching calls in the Clemson semi-final game (high probability).
Many of you have probably assumed that the College Football Playoff (CFP) Committee is inconsistent week-to-week or year-to-year on the criteria they evaluate to determine who is or isn’t in the playoffs. If I need to provide evidence, consider anytime Clemson or Oklahoma make it in with a weak schedule, or the time that the Alabama Crimson Tide did not play in a conference championship and jumped the Big 10 Champion.
The 13 member CFP Committee is represented with delegation from different conferences, independent civilians and sometimes former military members, all of which are rotated-out every three years.
It was my initial hypothesis going into this analysis that there were inherent biases, since there are always member affiliations or preferences beyond the team where they have to recuse themselves in the voting process (e.g., team preferences, conference affiliations, regionality, recency biases, etc.).
The voting process, to my understanding, is sort of like trying to guess what is behind each door on a game show, in that it takes place behind closed doors and in secrecy. Even if you guess correctly that it is a new car you don’t know if it is McLaren P1 or a Trabant.
Lace up your football cleats and strap on your helmet, you no longer need to postulate or theorize, because while I do have a day job building data models for clients (weather impact, hurricane, marketing impulse shock, etc.), I get bored and use my powers for good -- analyzing sports related topics.
I too have spent hours sobbing into my hands in the corner of my house when Ohio State is not deservingly picked on selection Sunday for the playoffs, which is why I decided to build a data model to analyze the criteria that the CFP Committee was using and to what extent.
.
.
.
continued