Use the following link for access to an Excel workbook with actual RPI ratings and ranks and other data for teams based on games through September 25, 2022: 2022 RPI Report 9.25.22. On the left of the RPI Report sheet, there are five color coded columns. These columns are based on the seasons from 2007 to the present (excluding the 2020 Covid-constrained season). They show the rank ranges as of this stage of the season from which #1 through #4 seeds have come. They also show the at large bubble range and that teams ranked #10 or better as of this stage of the season always have gotten at large selections.
For understanding the limitations of mathematical rating systems, it is worth paying attention to the color coded columns as they evolve through the course of the season. For this week, these are the rank ranges for seeds and at large selections:
#1 seeds: ranked #28 or better
#2 seeds: ranked #50 or better
#3 seeds: ranked #85 or better
#4 seeds: ranked #85 or better
At large selections:
Assured: ranked #10 or better
Bubble: ranked #11 through #148
When the regular season, including conference tournaments, is over, here are the ranges:
#1 seeds: ranked #7 or better
#2 seeds: ranked #14 or better
#3 seeds: ranked #23 or better
#4 seeds: ranked #26 or better
At large selections:
Assured: ranked #30 or better
Bubble: ranked#31 through #57
Week by week, the ranges will shrink from the current ones to the final ones. As you can see, at this week in the season, the RPI is not close to where it needs to be to be useful. This is a good reminder that the RPI is intended to be used for actual decisions only at once point in the season: the time when the Committee makes its at large selections and seeds.
The end-of-season ranges also are an indicator of how the Committee sees the accuracy of the RPI.
Seed Range. The range from which the Committee has picked seeds extends to rank #26. Since there are 16 seeds, this suggests that the Committee thinks (consciously or not) that in that rank range, the RPI can be up to 10 positions off in where it places teams.
At Large Selection Range. Teams ranked #30 or better always have gotten at large selections. The range from which the Committee has made the remaining at large picks extends to rank #57. The Committee picks a total of 33 at large teams. Among the top 33 teams, there is a reasonable chance there will be 7 to 9 automatic qualifiers (the Power 5 conference champions and the American, Big East, Ivy, and West Coast conference champions). This means the Committee will be trying to identify roughly the 40 to 42 best performing teams as the ones to get the 33 at large positions, given that about 7 to 9 of them will be automatic qualifiers. This suggests a number of things:
1. In the #30 or better rank range, the Committee thinks (consciously or not) that the RPI can be only up to 10 positions off in where it places teams. Put differently, if a team is ranked #30 by the RPI, it must be at least among the top 40 best performing teams.
2. In the #31 plus a little rank range, the Committee thinks (consciously or not) that the RPI can be off by 10 or more positions.
3. As the ranks get poorer, the Committee thinks (consciously or not) that the RPI can be off by increasing numbers of positions, including by as many as about 17 positions when it gets to the #57 ranked team.
4. Putting 2 and 3 together, in the #31 to #57 rank range, roughly speaking, an RPI rank difference of about 26 positions may not be sufficient to establish that the better ranked team has been the better performing team.
NOTE: The rank ranges are based on years during which tie games went to overtime, whereas starting this year there will be no overtime games except during conference tournaments. I do not believe the rule change will lead to a change in the rank ranges, but it could.
