Each year, eight teams will make the Women's Frozen Four. While four automatic bids will be awarded to the conferences (WCHA, Hockey East, ECAC, and CHA), the remaining four slots will be determined by what is called pairwise rankings. For information on the selection criteria, please visit this page at USCHO.com. An important component to the pairwise rankings is the Ratings Percentage Index (RPI). RPI is based on a formula, described below, that uses arbitrary weights to determine a teams rankings. A number of alternative methods have been suggested. The ratings here are based on what David Mease uses to rank college football teams. A major difference between college hockey and college football is that ties are possible. For these rankings, the probability of a tie is explicitly included in the model. For a detailed description of the model, please see the slides of a talk I gave at an Allegheny Mountain sectional meeting of MAA. In addition, A Bayesian approach has been adopted, with prior rating values based on last years results. For information of the prior distributions, click here.
These ratings are updated each morning. Data used to determine the rankings are non-exhibition games against Division I teams. Sacred Heart is not considered a D-I team for these calculations. Only the win/loss/tie result is used. Margin of victory and and location of game information are are not used.
As noted in the presentation above, the probability of two teams
is calculated by the algorithm. Currently, the estimated value of phi
approximately 0.195. Therefore, the probability of a tie:
For teams with equal ratings: 15.4%
For teams with a difference of 1 in their rating: 9.4%
For teams with a difference of 2 in their ratings: 2.1%
Detailed current D-I RPI can be found here. Quoting from USCHO.com, RPI uses only results from games between two teams that each play 20 or more games against Division I opponents. Factors involved are 1) the team's winning percentage; 2) the average winning percentage of the team's opponents; and 3) the average winning percentage of the team's opponents' opponents. These factors are multiplied by 30%, 24%, and 46% respectively. Please note that these rankings include all Division I teams and I will not being eliminating games against the weakest four opponents as the NCAA may or may not do.
My name is Michael Rutter and I am an Assistant Professor of Statistics and Mathematics at Penn State Erie-The Behrend College. I can be reached at firstname.lastname@example.org. I used to work at Mercyhurst College and I am currently an off-ice statistician for the Mercyhurst Lakers Women's Hockey team. I am known as “LakersFan” on the USCHO.com message boards, so feel free to comment there as appropriate.
This work is licensed under a Creative Commons License.