The Pairwise Rankings (PWR) are a statistical tool designed to approximate the process by which the NCAA selection committee decides which 12 teams to invite to the Divisions I championship tournament. Although the PWR does not precisely duplicate the method used by the committee, the PWR has, in actuality, exactly predicted the NCAA tournament entries in each of the last four years. (The difference in the process, is that the NCAA committee doesn't actually take the final step of totaling the comparison wins, and summarizing it into a neat PWR chart.
Note: The committee, in 1998, adopted language which effectively removes teams playing schedules of limited scope, such as the various MAAC programs, from full PWR consideration. These schools still appear in the PWR, but their rankings may not carry the same weight in tourney selection that they otherwise would.
The PWR method compares every team which owns a .500 or better overall record (a "team under consideration," or TUC) with every other such team, with the winner of each comparison earning one PWR point. After all possible comparisons are made, the points are totaled up and rankings listed accordingly.
For instance, if there are 24 TUCs, the greatest number of PWR points any one team could earn would be 23, by winning the comparison with each of the other 23 teams. Meanwhile, a team which lost all of its comparisons would, of course, have no PWR points.
Teams are then ranked by PWR point total, with ties broken by looking at the Ratings Percentage Index (RPI). Note: this tiebreaking procedure is used solely for convenience in displaying the PWR, and will not necessarily match the committee's process. This is especially true near the end of the top 12, where the committee looks more closely at head-to-head comparisons when selecting the last few teams. However, the committee does indeed use RPI to break ties within those head-to-head matchups. For example, if that head-to-head comparison is tied, or if there is a transitive tie in a three-way comparison (A def. B -- B def. C -- C def. A), then RPI is indeed used to break the deadlock.
In accordance with NCAA rules, only Division I programs playing a full D-I schedule are eligible for D-I tournament play, and hence only these schools are included in the PWR or in the RPI. Division II schools, irrespective of their conference affiliations or schedules, are ineligible -- this means, for instance, Bentley and Mercyhurst, though they are MAAC members, are not eligible for the D-I tourney. In addition, games played against ineligible teams are not included in RPI and PWR calculations.
So how are teams compared? The PWR uses five criteria which are combined to make a comparison: RPI rating, record in the last 16 games, record against all other TUCs, record against common opponents and head-to-head competition.
For an example, let's consider a hypothetical comparison between Michigan and Michigan State which might look like this:
Michigan vs Michigan State RPI 0.5891 0 0.5933 1 L16 11- 5- 0 0 12- 4- 0 1 TUC 8- 4- 1 0 8- 1- 1 1 H2H 0 2 COP 16- 1- 1 1 12- 2- 2 0 ============================================ PTS 1 5
Michigan State has the higher RPI, and the Spartans also have the better winning percentage in their last 16 games (12-4-0, .750 vs. Michigan's 11-5-0, .688) and against other TUCs (8-1-1, .850 vs. Michigan's 8-4-1, .631). Against common opponents -- teams both schools have played this season -- Michigan has the edge (16-1-1, .909 vs. Michigan State's 12-2-2, .813).
Each of the above is worth one point, so Michigan State leads Michigan 3-1 in these four criteria. Lastly, we throw in head-to-head competition, for which one comparison point is awarded for each win. Since Michigan and Michigan State have played two games this year and the Spartans have won both of them, Michigan State gets two more comparison points and the Wolverines get none.
Thus, Michigan State wins this comparison by the score of 5-1, and gets one PWR point. Notice that the final score of the comparison itself doesn't matter -- Michigan State only gets one PWR point no matter what the score of the comparison itself is.