RPI: Modified RPI?

Updated June 2025

Many, including some members of the Women's Soccer Committee, have wanted to use a different rating system than the NCAA RPI, or at least an additional rating system.

In 2023, the NCAA agreed to allow the Committee to use an additional rating system as one of several secondary criteria, in the event the Committee was not able to make its at large selection and seeding decisions based only on the primary criteria. The additional system was the KP Index, commonly known as the KPI, which is a rating system created by Kevin Pauga, an Associate Athletic Director at Michigan State.

In addition, I have created a modified version of the RPI, the Balanced RPI. And, Kenneth Massey has a rating system. The Committee is aware of both systems.

DESCRIPTION OF AND RATIONALE BEHIND THE BALANCED RPI

As shown on the RPI: Formula page, the NCAA RPI formula for Team A combines three elements with effective weights as follows:

Winning Percentage (WP): 50%

Opponents' Winning Percentages (OWP): 40%

Opponents' Opponents' Winning Percentages (OOWP): 10%

Within the formula, the OWP and OOWP portions together represent Team A's Strength of Schedule (SoS).

If you think of Team A as an opponent, its SoS contribution to each team it plays is its own winning percentage, which becomes part of OWP for the team it has played, and its opponents' winning percentages, which become part of OOWP for the team it has played. Thus under the NCAA RPI, Team A's SoS contribution to an opponent combines only two elements with effective weights as follows:

OWP: 40% x 2 = 80%

OOWP: 10% x 2 = 20%

As you can see, Team A's strength under the NCAA RPI is measured using three elements with 50%-40%-10% effective weights, whereas its strength as a SoS contributor is measured using only two elements with 80%-20% effective weights. As shown on the RPI: Measuring the Correlation ... page linked above, an effect of this is that there is a significant disconnect between teams' NCAA RPI ranks and their ranks under the NCAA RPI formula as strength of schedule contributors. One unfortunate result of this is that it enables and encourages teams with NCAA Tournament hopes to do their best to "trick" the NCAA RPI when doing their non-conference scheduling, so as to maximize their RPI ratings and ranks. And as shown on the RPI: Measuring the Correlation ... page, it causes the NCAA RPI to discriminate against teams from some conferences and regions and in favor of others.

I developed the Balanced RPI primarily to fix the NCAA RPI strength of schedule problem so that teams' RPI ranks and their ranks of Strength of Schedule contributors match. My objective was to end up with a formula that achieved this so that teams no longer would be able, or be encouraged, to "trick" the RPI through "smart" scheduling.

My basic approach in developing the Balanced RPI was as follows:

First, I wanted to get away from the NCAA RPI's emphasis, in the Strength of Schedule part of its formula, on an opponent's winning percentage as distinguised from the opponent's opponents' winning percentages. So in the initial RPI computation, I set the formula so that a team's winning percentage would have a 50% effective weight, the team's opponents' winning percentages would have a 25% effective weight, and the team's opponents' opponents' winning percentages would have a 25% effective weight.

With this change, there still are significant differences between teams' RPI ranks and their ranks as Strength of Schedule contributors to their opponents' ratings.

Second, with this new formula's ratings in hand, I wanted to get teams' RPI ranks and their ranks as Strength of Schedule contributors to come as close as possible to matching. To accomplish this, I did an additional computation. The formula for this computation had two elements, each weighted at 50%: (1) a team's winning percentage using the NCAA RPI's formula for computing winning percentage, and (2) a team's strength of schedule, computed as the average of the team's opponents' RPI ratings as computed in the first step. The theoretical basis for this was as follows:

In the initial RPI formula, the RPI rating and rank it produces, by NCAA definition, is a better measure of a team's strength than the Strength of Schedule portion of the formula. Thus, by definition, it is better to supplement the initial calculation with an additional one that uses teams' initial RPIs as their strength of schedule contributions to their opponents.

When I did this additional computation, it indeed did lessen the gaps between teams' rating ranks and their ranks as strength of schedule contributors, but it still left significant gaps.

I then did successive similar computations, each time using teams' winning percentages as the first element of their ratings and the average of their opponents' previous-step RPI ratings as the second element, with each element weighted at 50%.

Altogether, following the initial computation I did 14 additional rounds to end up with the Balanced RPI. At that point, teams' Balanced RPI ranks and teams' ranks as strength of schedule contributors were essentially the same.

The complete details of how to compute the Balanced RPI are at the bottom of this page.

COMPARING HOW THE NCAA RPI, THE KPI, THE BALANCED RPI, AND MASSEY PERFORM AS RATING SYSTEMS

With that background, the purpose of this page is to compare how the NCAA RPI, the KPI, the Balanced RPI, and Massey perform as rating systems.

To do the comparisons, I use my Correlator program, which performs a number of tests of how rating systems' perform. I explain those tests in detail at the RPI: Measuring the Correlation Between Teams' Performance and Their Ratings page. Rather than my re-explaining the tests here, I recommend you open the linked page in a separate window and for each test go to that page's explanation and then return to this page to see how that test evaluates each of the four systems. I will follow the same order of tests as at the RPI: Measuring the Correlation ... page. Also, on that page and this one, I use the same test headings and have highlighted the same headings with the same color. So if you scroll down this page to a particular heading, you then can scroll down to the same heading at the RPI: Measuring the Correlation ... page to find more expanded information about the test. In most cases at the RPI: Measuring the Correlation ... page, just above the heading you will find a Preliminary Calculations section that explains the data, calculations, and reasoning underlying the test.

Before going through the tests, here are some cautions to bear in mind:

NCAA RPI and Balanced RPI

For the NCAA RPI and Balanced RPI, I use ratings from 2010 through 2024, with game results from 2010 through 2021 adjusted as though the No Overtime rule had been in effect during those years. For the NCAA RPI, the ratings are as though the NCAA's 2024 RPI formula had been in effect for all years, including the 2024 bonus and penalty adjustments. For the Balanced RPI, I compute teams' winning percentages as though the NCAA's 2024 formula had been in effect. The Balanced RPI does not use bonus and penalty adjustments.

KPI

For the KPI, ratings are available only for the years since 2017. In addition, the ratings from 2017 through 2021 are based on games with overtimes allowed. Thus comparisons of the KPI's performance to that of the other systems is not an exact apples-to-apples comparison. In addition, for the Correlator to work it needs ratings to enough decimal places that no two teams have identical ratings. To accomplish this, for the KPI ratings I have implemented a fix (which is inconsequential in terms of how the ratings indicate team strength) so that all teams have unique ratings.

Massey

For Massey, for the years from 2010 through 2021, the ratings are based on games with overtimes allowed. Thus comparisons of Massey's performane to that of the other systems is not an exact apples-to-apples comparison. In addition, I have implemented a fix like that for the KPI so that all teams have unique ratings. And in addition, a number of years ago Massey made changes to the numerical range of his ratings and I have implemented a fix so that his ratings ranges are consistent across the years. Because of this latter change, I consider the Correlator test results for Massey very slightly less reliable than for the other systems, although sufficiently reliable for purposes of comparing how his system works as compared to the other systems.

With that background, here is how the four rating systems' performances compare.

ALL TEAMS AS A GROUP

In this table, a difference of 0.1% in the first three percentage columns represents about 3 games per year out of about 3,000 games. As you can see, the differences among the systems are not large.

TOP 60 TEAMS

In this table, a difference of 0.1% in the first three percentage columns represents about 1 game per year out of about 1,000 games. Since the number of ties varies from one system to another depending on which teams a system has in the Top 60, the best column to look at for comparisons is the Disregarding Ties column. As you can see, the differences among the systems are not large.

INDIVIDUAL TEAMS, ACTUAL LESS LIKELY WINNING PERCENTAGE DIFFERENCES

As this table shows, the NCAA RPI performs better than the KPI in terms of teams' expected results (based on their ratings) being consistent with their actual results. The Balanced RPI and Massey perform better than the NCAA RPI and KPI, with the Balanced RPI performing the best.

INDIVIDUAL TEAMS, RATING SYSTEM RANK v SYSTEM STRENGTH OF SCHEDULE CONTRIBUTOR RANK DIFFERENCES

In this table, it is not possible to compare the KPI or Massey systems' ranks of teams as compared to their strength of schedule contributor ranks, since those systems do not report teams' ranks as strength of schedule contributors.

As you can see from the table, however, the NCAA RPI has large differences between teams' RPI ranks and their ranks as strength of schedule contributors. Thus teams are able, and encouraged, to "trick" it through smart scheduling.

On the other hand, for the Balanced RPI, teams' ranks and their strength of schedule contributor ranks are essentially the same so it is not possible for teams to "trick" it. This means that under the Balanced RPI, teams would be able to take a "what you see is what you get" approach to scheduling. For example, if a team believes a potential opponent will be ranked #40, the team also can expect the rating system will give it credit for playing the #40 opponent.

CONFERENCES, OVERALL PERFORMANCE FOR CONFERENCES

This table shows that for the Balanced RPI and Massey, conference teams' actual results are far more consistent with their ratings than for the NCAA RPI and the KPI. Massey also performs better than the Balanced RPI.

CONFERENCES, RELATIONSHIP BETWEEN CONFERENCE PERFORMANCE AND CONFERENCE RATING

Here are four charts for the four rating systems, in the order NCAA RPI, KPI, Balanced RPI, Massey:

The following table summarizes what the four charts show:

As this table shows, both the NCAA RPI and the KPI discriminate against stronger conferences and in favor of weaker ones. On the other hand, the Balanced RPI and Massey do not discriminate. Further, if you look at the R-squared values on the charts, you will see that for the NCAA RPI and the KPI, the conference performance distributions in relation to conference strength are not random but rather have a fairly high correlation to conference strength. For the Balanced RPI and Massey, however, the conference perfomrance distributions in relation to conference strength are virtually random. Thus the Balanced RPI and Massey have solved the problem of discrimination in relation to conference strength.

CONFERENCES, RELATIONSHIP BETWEEN CONFERENCE PERFORMANCE AND CONFERENCE PARITY AS INDICATED BY PROPORTION OF IN-CONFERENCE TIES

Here are four more charts for the four rating systems, in the order NCAA RPI, KPI, Balanced RPI, Massey:

The following table summarizes what the four charts show:

Using the percentage of in-conference ties as a surrogate for in-conference parity, this table suggests that the NCAA RPI discriminates the most against conferences with high in-conference parity. The KPI has less discrimination and the Balanced RPI even less. Massey has minimal discrimination. The R-squared values indicate that the correlation between in-conference parity and conference performance in relation to ratings is low for the NCAA RPI and KPI and fairly low for the Balanced RPI. In other words, in-conference parity may play a role in the rating systems' discrimination among conferences, but the role is relatively small; and in any event, the order of discrimination is the NCAA RPI with the most, followed by the KPI, then the Balanced RPI, and then Massey with almost no discrimination.

CONFERENCES, RELATIONSHIP BETWEEN CONFERENCE PERFORMANCE AND THE DIFFERENCE BETWEEN CONFERENCES' RATINGS AND THEIR RATINGS AS STRENGTH OF SCHEDULE CONTRIBUTORS

Here are two charts, in the order NCAA RPI and Balanced RPI. There are no charts for the KPI or Massey because strength of schedule contributor ratings are not available for either of them.

The following table summarizes these charts:

As this table shows, for the NCAA RPI, conferences that are underrated (perform better than their ratings say they should) have higher differences between their RPI ratings and their strength of schedule contributor ratings; and conferences that are overrated have lower differences. Further, the high R-squared value on the NCAA RPI chart shows a strong correlation between conferences' performance and the extent of the difference between their RPI ratings and their strength of schedule contributor ratings.

As discussed on the RPI: Measuring the Correlation ... page, careful thinking suggests that the way the NCAA RPI computes strength of schedule will cause the NCAA RPI to discriminate against stronger conferences and in favor of weaker ones and also to discriminate against conferences with high in-conference parity and in favor of those with low in-conference parity. The above charts for the NCAA RPI confirm this is the case.

On the other hand, the Balanced RPI shows no such discrimination. And, the low R-squared value on its chart indicates there is no connection between conferences' performance and the differences between conferences' Balanced RPI ratings and strength of schedule contributor ratings. Thus it is clear that the Balanced RPI has solved the NCAA RPI's problem.

REGIONS, OVERALL PERFORMANCE FOR REGIONS

This table shows that for the Balanced RPI and Massey, region teams' actual results are far more consistent with their ratings than for the NCAA RPI and the KPI. As between the Balanced RPI and Massey, Massey performs better.

REGIONS, RELATIONSHIP BETWEEN REGION PERFORMANCE AND REGION RATING

Here are four charts for the four rating systems, in the order NCAA RPI, KPI, Balanced RPI, Massey:

The following table summarizes what the four charts show:

As this table shows, the NCAA RPI and KPI discriminate among regions in relation to region strength. The Balanced RPI and Massey do not discriminate among regions in relation to region strength. Further, the R-squared value on the charts indicate that for the NCAA RPI and KPI, especially the NCAA RPI, the relationship between the discrimination and region strength is not random. For Massey, the R-squared value may suggest that region performance is not completely independent of region strength, but as stated, in effect Massey does not discriminate in relation to region strength, so its low-moderate R-squared value appears to be fortuitous. For the Balanced RPI, the R-squared value indicates there is no relationship between region performance and region strength.

REGIONS, RELATIONSHIP BETWEEN REGION PERFORMANCE AND REGION PARITY AS INDICATED BY PROPORTION OF IN-REGION TIES

Here are four more charts for the four rating systems, in the order NCAA RPI, KPI, Balanced RPI, Massey:

The following table summarizes what the four charts show:

As this table shows, the NCAA RPI and KPI discriminate among regions in relation to in-region parity, as measured by the proportion of in-region ties. The Balanced RPI has only a little discrimination among regions in relation to in-region parity. Massey has virtually no discrimination. The R-squared values for the NCAA RPI and KPI suggest the relationship between in-region parity and region performance in relation to ratings is not random. The R-squared value for the Balanced RPI indicates a strong relationship between in-region parity and region performance in relation to ratings, but as stated the Balanced RPI shows only a little discrimination. Even Massey has an R-squared value indicating a relationship between in-region parity and region performance in relation to ratings, but again as stated Massey has virtually no discrimination.

REGIONS, RELATIONSHIP BETWEEN REGION PERFORMANCE AND THE DIFFERENCE BETWEEN REGIONS' RATINGS AND THEIR RATINGS AS STRENGTH OF SCHEDULE CONTRIBUTORS

Here are two charts, in the order NCAA RPI and Balanced RPI. There are no charts for the KPI or Massey because strength of schedule contributor ratings are not available for either of them.

The following table summarizes what the two charts show:

As this table shows, for the NCAA RPI, regions that are underrated (perform better than their ratings say they should) have higher differences between their RPI ratings and their strength of schedule contributor ratings; and regions that are overrated have lower differences. Further, the high R-squared value of roughly 0.93 on the NCAA RPI chart indicates the trend line has a strong connection to the data, in fact suggesting that the differences between regions' RPI ratings and their strength of schedule contributor ratings are the cause of the NCAA RPI's discrimination.

On the other hand, the Balanced RPI shows little discrimination and -- through its R-squared value -- a moderately strong connection between regions' performance and the differences between regions' RPI ratings and strength of schedule contributor ratings.

CONCLUSIONS ABOUT THE RATING SYSTEMS

Regardless of the rating system, there always will be some game results that are inconsistent with the opponents' ratings. Given the relatively few games that individual teams play, one should expect that different teams will have different proportions of game results that are inconsistent with their ratings. When looking at the different conferences' teams bunched together, where there are many more games per conference, one still should expect different conferences to have different proportions of game results that are inconsistent with ratings but the differences should be smaller. And when looking at the different regions' teams bunched together, where there are even more games per region, one still should expect different regions to have different proportions of game results that are inconsistent with ratings but the differences should be even smaller. In an ideal system, however, the distribution of the proportions of inconsistent results, whether for individual teams, for conferences, or for regions will be random.

A critical question, for a rating system, is whether the distributions of the proportions of inconsistent results are part of the expected, inevitable randomness or whether they have non-random patterns and thus are due to a flaw in the rating system.

As the information on this page shows, there is a clear difference between how the NCAA RPI and KPI perform as rating systems and how the Balanced RPI and Massey perform. The NCAA RPI and KPI have patterns of discrimination against stronger conferences and regions and against conferences and regions with higher in-conference and in-region parity. For the NCAA RPI, its patterns of discrimination are tied to the way it computes strength of schedule. It is not possible to determine why the KPI has these patterns, due to the lack of information about the KPI's inner workings, we only know that the patterns are similar to the NCAA RPI's.

The Balanced RPI and Massey either do not have these patterns of discrimination or have them at a much lower level than for the NCAA RPI and KPI.

Because of the differences between how the NCAA RPI and KPI perform, on the one hand, and how the Balanced RPI and Massey perform, on the other hand, we know that the NCAA RPI's and KPI's distributions of the proportions of inconsistent results are not inevitable. Instead, they are due to flaws in the NCAA RPI and KPI rating systems. Further, we know that the flaws in the NCAA RPI could be mostly or entirely avoided while still using the NCAA RPI's underlying data and core structure, through the Balanced RPI's modifications to the NCAA RPI formula.

THE COMMITTEE HAS ADDED THE KPI AS A SUPPLEMENT TO THE NCAA RPI. WHY THE KPI AND NOT MASSEY OR THE BALANCED RPI?

Beginning with the 2023 season, the NCAA allowed the Women's Soccer Committee to use the KPI as a secondary criterion during the NCAA Tournament at large selection and seeding process.

The decision to allow use of the KPI came following a recommendation from the Committee at its January 30-31, 2023 meeting:

"Action Item. ... 2. Nonlegislative item. *KPI.

"a. Recommendation. That the KPI be used as another consideration during ranking and selection.

"....

"c. Rationale. The KPI is an established tool used that has been proven effective in basketball selections. After the presentations from Kevin Pauga at last year's annual meeting and a review of the KPI from 2022 at this meeting, the committee feels that having another metric to consider in comparing teams is helpful. The KPI data was similar to the RPI data; however, instances where it differed prompted the committee to more closely compare those teams, continue discussion, and identify and discern imputs that contributed to the differences.

"Like the RPI, the committee would use the KPI as a selection tool. It also will be a valuable addition when the committee discerns teams' resumes during the seeding process."

Report of the NCAA Division I Women's Soccer Committee January 30-31, 2023, Meeting.

Following the 2023 season, however, the Committee changed its opinion on use of the KPI and recommended use of Massey instead:

"Informational Items. ... 6. Review of the 2023 championship. c. Committee ranking calls and selection weekend.

(4) KPI, and other ranking systems. The committee received information from Ken Massey on what the Massey Ratings entail in addition to receiving information from Chris Thomas on his ranking metric. The committee felt that KPI did not add much help in the selection room and preferred to review a predictive-based metric such as the Massey Ratings and is therefore requesting the ability to review them during the year with the likelihood of proposing the ability to use the rating for the 2025 selections."

Report of the NCAA Division I Women's Soccer Committee January 29-30, 2024 Meeting.

But following the 2024 season, the Committee decided to stay with the KPI:

"Informational Items. ... 5. Review of the 2024 championship. c. Committee ranking calls and selection weekend.

(3) KPI, and other ranking systems. The committee felt that the KPI should be used more in selections as a valued tool in the process. The committee reviewed the Massey Ratings and decided not to request it as an additional selection criterion at this point."

Report of the NCAA Division I Women's Soccer Committee December 9, 2024, and January 29, 2025, Meetings.

With this history, the question is: Why did the Committee decide to continue using the KPI rather than using Massey (or the Balanced RPI)?

As stated in the Report from the January 2023 meeting, the Committee in 2022 had met with Kevin Pauga about the KPI system. I know that some Committee members had wanted to look at a rating system outside the NCAA RPI. I do not know why the KPI was the system they heard about in 2022, but a Committee member told me the Committee did not select the KPI as the system to look at. So, I suspect that in response to the Committee wanting to look at an outside system, the NCAA staff arranged for the Committee to have a look at the KPI.

Also according to the Report from the January 2023 meeting, after the 2022 season the Committee members reviewed the KPI rankings for the 2022 season. The Committee found that the KPI rankings were similar to the NCAA RPI rankings, but in some cases were different. The Committee then was able to look more closely at the teams involved in the differences to see what underlying data contributed to the differences. Presumably, the Committee felt this was something they would be able to do during upcoming NCAA Tournament at large selection and seeding, as a way of improving the process.

The following table is from the 2022 season. It shows the Top 57 teams under the NCAA RPI and the Top 57 under the KPI. The reason for using the Top 57 is that historically, no team ranked poorer than #57 has gotten an at large selection to the NCAA Tournament. Thus for practical purposes, the Top 57 has all the legitimate candidate teams for a Tournament at large position.

In the table, the green highlighting is for teams in the NCAA RPI's Top 57. The salmon is for the KPI's Top 57. The NCAA Seed or Selection column shows the Committee's seed and at large selection decisions and also unseeded automatic qualifiers. The #1 to #4 seeds are represented by 1 through 4; the #5 through #8 seeds by 4.5 through 4.8; unseeded automatic qualifiers by 5; unseeded at large selections by 6; and teams not getting at large selections by 7.

As you can see, there were three teams in the NCAA RPI's Top 57 that were outside the KPI's:

Auburn (NCAA RPI #50, KPI #60)

Northeastern (#56/#58)

South Dakota State (#57/#69)

As you also can see, none of these got an at large selection.

The three teams in the KPI's Top 57 but not in the NCAA RPI's were:

Nebraska (58/55)

Gonzaga (59/52)

Mississippi (60/57)

For these six "affected" teams, it is notable that the differences between the two systems' ratings were not large. Further, there was only a three team difference. In other words, the differences between the NCAA RPI and KPI were small. This would have made it relatively easy for the Committee, in its review process, to compare the detailed profiles of Nebraska, Gonzaga, and Mississipi to the profiles of the last three teams that actually got 2022 at large selections, to see if adding the KPI as a resource might have caused the Committee to make different selections. In other words, (1) the differences between the NCAA RPI and KPI ratings were small enough to not create significant controversy among those with interests in the selections and (2) the differences were manageable as part of the Committee process.

In 2023, the Committee actually got to use the KPI, as a secondary criterion. Although the secondary criteria are to be used only if the Committee is not able to make its decisions using the primary criteria, the Committee's decisions over the years indicate that the Committee always uses both the primary and secondary criteria. Here is what the Committee saw in 2023:

Again in 2023, there was only a three team difference between the NCAA RPI's Top 57 and the KPI's. The three in the NCAA RPI's Top 57 but not in the KPI's were:

Lamar (51/81) Automatic Qualifier

Duke (53/62) No at large selection

Samford (56/63) No at large selection

The three in the KPI's Top 57 but not in the NCAA RPI's were:

Bucknell (59/54) Automatic Qualifier

Kentucky (60/53)

Wake Forest (82/52)

This year, there were two cases where the rank differences were pretty large: Lamar dropping 30 rank positions and Wake Forest rising 30 positions. Lamar was an automatic qualifier so did not matter. KPI addition Bucknell was an automatic qualifier. With Duke and Samford not having gotten at large selections, this would have left the question whether Kentucky and/or Wake Forest should get an at large selection as compared to the last teams that actually got them.

For this year, the difference in how the systems ranked Wake Forest was pretty large, which might make one wonder about the NCAA RPI as a rating system. On the other hand, the number of teams involved was small enough to be easily manageable.

Nevertheless, in its evaluation of using the KPI following the 2023 season, the Committee concluded that the KPI did not add much to the process and asked to be able to use Massey instead starting with the 2025 season. This left the Committee still using the KPI for the 2024 season, but knowing it might be using Massey beginning in 2025.

Here is what the Committee saw in the 2024 season, using the KPI:

This year, there were four team changes in going from the NCAA RPI to the KPI. The teams in the NCAA RPI's Top 57 but not in the KPI's were:

California (49/58) At large selection

Boise State (52/64) No at large selection

Dayton (53/59) No at large selection

SMU (56/78) No at large selection

The teams in the KPI's Top 57 but not in the NCAA RPI's were:

Buffalo (58/48)

Texas A&M (60/51) Disqualified from at large selection due to winning percentage below 0.500

Utah Valley (61/56)

Arizona (63/57)

As you can see, none of the rating differences was large. And, with only three new at large candidates to consider, the process of considering the KPI would have been manageable.

It seems reasonable to consider the possiblity that after the 2024 season, the Committee looked to see what the effect might have been if it had used Massey instead of the KPI. Here is what the Committee would have looked at:

As you can see, there is a 10-team shift between the NCAA RPI and Massey. Further, some of the rank differences are very large. Neither these differences nor the teams that, under Massey, drop out of or enter into the Top 57 are surprising if you consider them in the context of all the information above comparing how the rating systems perform.

The following teams were in the NCAA RPI's Top 57 but not in Massey's:

Western Michigan (20/61) Automatic Qualifier, Not seeded though well within seed candidate range

Fairfield (36/96) Automatic Qualifier, Not Seeded though within seed candidate range

Massachusetts (42/77) No at large selection

South Florida (44/67) No at large selection

Liberty (47/81) No at large selection

James Madison (48/62) Automatic Qualifier

Boise State (52/60) No at large selection

Dayton (53/70) No at large selection

Stony Brook (54/88) Automatic Qualifier

SMU (56/65) No at large selection

The following teams were in Massey's Top 57 but not in the NCAA RPI's:

Texas A&M (60/57) Disqualified from at large selection due to winning percentage below 0.500

Utah Valley (61/49)

Clemson (64/47) Disqualified from at large selection due to winning percentage below 0.500

Baylor (65/51)

Boston College (66/48)

Gonzaga (69/56)

Harvard (76/54)

Alabama (87/46)

Utah (90/55)

Arizona State (141/45)

You can see that if the Committee had used Massey as its additional rating system, (1) some of the differences between the NCAA RPI and Massey ranks would have been large enough to create significant controversy among those with interests in the selections, indeed, the differences might have called the validity of the NCAA RPI itself into question; and (2) there would have been enough differences to make the Committee process much more difficult to manage.

Alternatively, if the Committee in 2024 had considered the Balanced RPI, here is what it would have seen:

As you can see, there is an 8-team shift between the NCAA RPI and the Balanced RPI. Further, some of the rank differences are large, though not as large as the largest for Massey. Neither these differences nor the teams that, under the Balanced RPI, drop out of or enter into the Top 57 are surprising if you consider them in the context of all the information above comparing how the rating systems perform.

The following teams were in the NCAA RPI's Top 57 but not in the Balanced RPI's:

Fairfield (36/81)

Massachusetts (42/58)

South Florida (44/65)

Liberty (47/72)

James Madison (48/63)

Dayton (53/64)

StonyBrook (54/84)

SMU (56/68)

The following were in the Balanced RPI's Top 57, but not in the NCAA RPI's:

Utah Valley (64/51)

Arizona (63/43)

Baylor (65/51)

Boston College (66/55)

Gonzaga (69/53)

Alabama (87/57)

Loyola Marymount (88/50)

Illinois (94/52)

Again you can see as with Massey that if the Committee had used the Balanced RPI as its additional rating system, (1) some of the differences between the NCAA RPI and Balanced RPI ranks would have been large enough to create significant controversy among those with interests in the selections, indeed, the differences might have called the validity of the NCAA RPI itself into question; and (2) there would have been enough differences to make the Committee process much more difficult to manage.

So, to return to the question: Why the KPI and not Massey or the Balanced RPI?

In the financial world, there is something known as the Sunk Cost Fallacy. This is reasoning that one should continue with an activity because of the resources already invested in it. It also is known as "throwing good money after bad."

And, there is an adage known as Segal's Law: A man with a watch knows what time it is. A man with two watches never is sure.

Although I do not know, it is possible that thought patterns like the Sunk Cost Fallacy and Segal's Law are at play in the Committee decision to continue using the KPI as its additional rating system rather than shifting to Massey or the Balanced RPI. The NCAA has used the NCAA RPI for years (despite knowing it has problems) and the Committee is used to it. The KPI numbers, as the Committee stated in its January 2023 report, are "similar" to the NCAA RPI. Thus adding the KPI as a Committee tool will not "make waves" or be unmanageable. On the other hand, using Massey or the Balanced RPI would bring into question the legitimacy of the NCAA RPI as a rating system and would make it much harder for the Committee to manage its decision-making process.

In any event, for now the Committee will continue to use ratings that unfairly discriminate among conference and regions as described above, notwithstanding that other nondiscriminatory systems are available.

*****************************************************************************************************************************************************

COMPUTATION OF THE BALANCED RPI

Step 1 - RPI 50 50 SoS

Compute the RPI just as the NCAA currently computes it (but with no bonus or penalty adjustments), but using the following formula:

(WP + (1.4*OWP – 0.0046) + (2.4*OOWP-0.0140))/4

In this formula, the 1.4 and 2.4 numbers are multipliers that put the effective weights of the three elements at 50-25-25. The -0.0046 and -0.0140 numbers simply are centering adjustments that keep the ratings within ranges we are used to seeing and have no effect on the resulting rankings. Each subsequent step will include a multiplier and a centering adjustment. For each step, the multiplier is set so that a team’s Winning Percentage and its Strength of Schedule each have a 50% weight.

Changing the ratios of the three RPI Elements is not unprecedented. The original RPI formula had different ratios than the current formula and ice hockey currently uses different ratios than the other sports (including different ratios for the men than for the women).

Step 2 - Iteration 2