RPI: Regional Issues

Updated June 2025

In addition to playing in "formal" conference pools, teams also tend to play within geographic regions.  Analyses of teams' schedules since 2007 show that teams tend to play within one of four geographic regions based on the states where they are located:  Middle, North, South, and West.  Here are the states whose schools are in each region:

Middle: Illinois, Indiana, Iowa, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, Wisconsin

North: Connecticut, Delaware, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont, Washington DC

South: Alabama, Arkansas, Florida, Georgia, Kansas, Kentucky, Louisiana, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, West Virginia

West: Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, Wyoming

The following table shows the number and percentage of games each region's teams have played against teams from their own and the other regions, since 2007:

DOES THE NCAA RPI HAVE A PROBLEM IN RELATION TO REGIONS AND, IF SO, HOW GREAT IS IT?

On the RPI: Measuring the Correlation Between Teams' Ratings and Their Performance page,  I showed in detail how my Correlator system can evaluate the performance of a rating system.  As an example, I showed how the Correlator evaluates the 2024 NCAA RPI formula that the NCAA currently uses for Division I women's soccer.  The material on that page includes an analysis of how well the NCAA RPI performs when rating teams from a region in relation to teams from other regions.  It shows that that this is an area where the NCAA RPI has a problem.

For details about the NCAA RPI's problem rating teams from a region in relation to teams from other regions, go to the linked page and scroll most of the way down to "Regions, Preliminary Calculations."  You will find the detailed information there and below, with tables and charts.

Part of the information under "Regions" relates to the impact of regional parity on ratings.  I suggest there that the NCAA RPI discriminates against teams from regions with high levels of parity and in favor of some teams from regions with low levels of parity (and provide similar information related to  in-conference parity).  Here is a more in depth explanation of why that is the case.

As I showed on the RPI: Formula page, the NCAA RPI formula gives a team's Winning Percentage a 50% effective weight, its Opponents' Winning Percentage a 40% effective weight, and its Opponents' Opponents' Winning Percentage a 10% effective weight.  Because of the high weight the NCAA RPI assigns to the Opponents' Winning Percentage element, it is much better from an NCAA RPI perspective to play opponents who will have good winning percentages even if against weak opponents than to play opponents who will have lesser winning percentages but against strong opponents.

Coupled with this is the fact that for financial and other reasons, almost all teams play most of their non-conference games within their own geographic regions.

With that as background, consider the following hypothetical:

There are two regions.   Each region, as expected, has a distribution of team strengths in a bell curve shape:  It has few teams at the weakest end of the rating spectrum, few teams at the strongest end, and the highest number of teams near the middle.  In the hypothetical, however, one region has a wide disparity of team strengths, distributed in a wide bell curve.  The other region has a high parity of team strengths, distributed in a narrow bell curve.

In non-conference scheduling, the bell curves are like food chains.  At the weak end of the bell curve, teams (Tier 1) are like the bottom of the food chain.  The next teams in the chain (Tier 2) can "feed" on them, building up their winning percentages.  Likewise the next teams (Tier 3) can "feed" on the Tier 1 and Tier 2 teams, building up their winning percentages.  And, as we progress across the bell curve, teams in successive tiers can "feed" on those teams in the weaker tiers, especially the ones with the best  winning records for their tiers.  This lets teams in the stronger tiers build up their winning percentages while also playing opponents who will have decent winning percentages.  And, it lets teams from the stronger tiers carry their own good winning percentages into conference play, where their good winning percentages will help boost their conference opponents' Opponents' Winning Percentages and Opponents' Opponents' Winning Percentages.

In the wide disparity region, what I described in the preceding paragraph works well so that teams on the strong side of the bell curve are able to build up their winning percentages against opponents with good winning percentages.  In the high parity region, however, it does not work well: Because of the parity, there are fewer teams able to achieve good winning percentages and thus fewer potential opponents likely to have good winning percentages.  In food chain terms, there are fewer good "feeding" opportunities.

Given the effective weights of the RPI elements, in the hypothetical one would expect the NCAA RPI to favor teams on the strong side of the wide disparity region's bell curve and to disfavor teams on the strong side of the high parity region's bell curve.

In that context, consider the following chart based on real data:

This chart shows the winning percentage distribution bell curves for each of the four regions: Middle (dark blue), North (red), South (green), and West (blue).  In the chart, the data points represent the numbers of teams the regions have at the different winning percentage levels.  (Since the regions have different numbers of teams, the numbers have been "equalized" to show the distributions as if the regions all had the same number of teams.)  The bell curves are computer generated to better show the distribution of teams in the regions.

There are two key things to notice about the chart:

The chart has the poorest winning percentage teams at the left and the best at the right.  Thus the bell curves more to the left are for regions with poorer winning percentages and those to the right are for regions with better winning percentages.  You can see that the West bell curve (blue) tends to be the most rightward.  Next comes the South (green), followed by the Middle (dark blue) and then the North (red).

The South (green) bell curve has the lowest apex and is relatively wide, which means the South has the fewest teams in the middle of the curve with more on either side, in other words a relatively high disparity of teams' winning percentages.  The West (blue) bell curve has the highest apex and is relatively narrow, which means the most teams in the middle of the curve with fewer on either side, in other words a relatively high parity of teams' winning percentages.  The Middle (dark blue) and North (red) have about the same apexes, between those of the West and South, but with the North's distribution of winning percentages overall more on the poor side than the Middle's and with the North's  disparity of teams on either side of the apex appearing to be greater than the Middle's..

With these region bell curves, based on the discussion of the hypothetical above, one would expect that the NCAA RPI would favor the South with its wide disparity bell curve and disfavor the West with its high parity bell curve.  One might also expect the NCAA RPI to slightly disfavor the Middle and to favor the North.

If you look at the first table under Regions, Preliminary Calculations at the RPI: Measuring the Correlation ... page, what they show is consistent with what one would expect based on the region winning percentage distribution curves.  This definitely is the case for the West and the South and seems likely so, though less clearly, for the Middle and North.

What this shows is that the NCAA RPI, because of the high weight it assigns to Opponents' Winning Percentage, cannot properly rate teams from different regions in relation to each other if some regions have a relatively high degree of winning percentage disparity and others have a relatively high degree of parity.  Rather, the NCAA RPI will favor teams from regions with the high degree of disparity and disfavor teams from regions with the high degree of parity.