Conference Finals
Using the power rankings that excluded the meaningless games (in my opinion), which are listed in last week's post, the probability of each of the remaining teams winning their conference and the Super Bowl, respectively, are listed below. I have a gut feeling I just butchered the grammer in the last sentence. Oh well . . . I'm a statistician - not a writer.
Probability win Conference / Prob win Super Bowl / Team
0.95647837 / 0.450289880 / Phi
0.04352163 / 0.002488641 / Atl
0.77729986 / 0.463053397 / Pit
0.22270014 / 0.084168082 / NE
The results from the Bradley-Terry logistic regression model are in log odds ratios. Basically, this means that the outcome of a game can be put into probability terms based on the performance of each team throughout the season, and the home-field advantage effect. The probability of each team winning the Super Bowl depends on the probability of winning their conference, as well as the conditional probabilities of beating either team in the other conference:
Probability the Patriots win the Super Bowl (P=probability):
P(Patriots beat Steelers in Pittsburgh)*P(Eagles beat Falcons in Philly)*P(Patriots beat Eagles)+P(Patriots beat Steelers in Pittsburgh)*P(Falcons beat Eagles in Philly)*P(Patriots beat Falcons)
It is assumed that there is no home-field advantage effect for the Super Bowl.
