RPI: Non-Conference RPI

Updated June 2025

INTRODUCTION

In addition to the NCAA RPI, the NCAA calculates a Non-Conference RPI (NCRPI).    As important background, on average, teams play 18.6 games per year (including conference tournament games), of which 10.4 are conference and 8.2 are non-conference.

The NCAA NCRPI and related information shows up in a number of reports the NCAA produces during the course of the season:

Summary Report, which includes:

Individual team's Adjusted NCRPI rank (but not rating)

For each team, its conference's  Adjusted RPI rating and rank (Note: This is not the conference's Adjusted NCRPI.)

Non-Conference Ranking Report, which includes:

Individual team's non-conference record (W-L-T) and non-conference winning percentage rank (but not the actual winning percentage)

Individual team's non-conference opponents' average winning percentage rank (but not the actual average winning percentage)

Individual team's non-conference opponents' opponents' average winning percentage rank (but not the actual average winning percentage)

Individual team's unadjusted NCRPI rank (but not rating)

Individual team's adjusted NCRPI rank (but not rating)

Conference Ranking Report (based on both conference and non-conference games), which includes:

Individual conference's record (W-L-T) and winning percentage rank (but not the actual winning percentage)

Individual conference's opponents' average winning percentage rank (but not the actual average winning percentage)

Individual conference's opponents' opponents' average winning percentage rank (but not the actual average winning percentage)

Individual conference's unadjusted RPI rank (but not rating)

Individual conference's adjusted RPI rank (but not rating)

Conference Non-Conference Ranking Report, which includes:

Individual conference's non -conference record (W-L-T) and winning percentage rank (but not the actual winning percentage)

Individual conference's non-conference opponents' average winning percentage rank (but not the actual average winning percentage)

Individual conference's non-conference opponents' opponents' average winning percentage rank (but not the actual average winning percentage)

Individual conference's non-conference unadjusted RPI rank (but not rating)

Individual conference's non-conference adjusted RPI rank (but not rating)

CALCULATION OF THE NCRPI

The NCAA computes individual teams' NCRPI ratings using the same basic format as for the NCAA RPI.  For information on the basic NCAA RPI format, go to the RPI: Formula page.  The details of the NCRPI computation for Team A are:

NCRPI Element 1:  This is Team A's winning percentage in non-conference games.  In other words, in computing Element 1, the NCAA disregards a team's conference games.  Beginning in 2024, for the regular NCAA RPI, Element 1 values a tie as 1/3 of a win whereas previously it valued a tie as 1/2 a win.  For the NCAA NCRPI, however, Element1 still values a tie as 1/2 a win.  (I do not know whether this is intentional or an oversight.)

NCRPI Element 2:  This is the average of Team A's non-conference opponents' winning percentages against teams other than Team A.  In identifying the non-conference opponents' winning percentages, the NCAA includes all the non-conference opponents' games, both conference and non-conference.

NCRPI Element 3:  This is the average of Team A's non-conference opponents' opponents' winning percentages.  In identifying the non-conference opponents' opponents' winning percentages, the NCAA includes all the non-conference opponents' opponents' games, both conference and non-conference.

NOTE: I believe the reason the NCAA includes, in Elements 2 and 3, opponents' conference and non-conference games, is to have additional games in the data base at least for those Elements even though the games are not included in Element 1.  Without agreeing that the NCRPI is a useful rating system with those games included, it is safe to assume it would not be useful without those games.

Unadjusted NCRPI:  Using data and NCAA published NCRPI rankings, I have validated that the computations of individual teams' three NCAA NCRPI elements are as indicated above.  I expected that the following formula, which matches the formula for computation of the regular NCAA RPI, would produce team NCAA NCRPI rankings that would match the NCAA's published NCRPI rankings:

NCRPI = (NCRPI Element 1 + 2*NCRPI Element 2 + NCRPI Element 3)/4

I expected the NCAA's and my NCRPI ranks to match because a number of years ago, one of the NCAA staff RPI experts wrote to me that the formula for computing the NCAA NCRPI is the same as for the regular NCAA RPI.  Using this formula for the NCRPI, however, does not produce rankings that match the NCAA's.

On investigation, I discovered that for some teams, the NCAA's published non-conference W-L-T numbers are higher than they should be.  As it turns out, for some conference tournament games, the NCAA's non-conference W-L-T numbers are counting the games as non-conference games.  This is true for each year for which the NCAA reports are currently available.

As a result of these NCAA errors,  the end-of-regular-season team NCAA NCRPI rankings are not correct,  The end-of-regular-season improper rankings, however, go beyond the teams whose conference tournament games were improperly identified.   This suggests to me that the problem is not simple data error entries for the improperly identified games.  Rather, based on a problem with the NCAA numbers several years ago (that the NCAA was able to correct), I suspect this problem with the NCAA NCRPI rankings has to do with the NCAA data not having been "sorted" properly before doing the final calculations that produce teams' NCRPI rankings.

The NCAA staff is aware of this problem with its NCRPI computations and intends to try to find and correct the problem.  In the meantime, I am assuming that the NCRPI formula is as stated above and that my NCRPI ratings and rankings are correct.

Bonus and Penalty Adjustments:  The NCAA uses an adjusted NCRPI.  It uses the same bonus and penalty adjustment structure and amounts that it uses for the adjusted NCAA RPI.  As it does in computing the bonuses and penalties that result in the adjusted NCAA RPI, the NCAA bases Team A's adjustment amounts on its opponents' unadjusted NCAA RPI rankings (and not on the opponents' unadjusted NCRPI rankings).  For information on the bonus and penalty structure and amounts, go to the RPI: Formula page.

Conference NCAA RPI and NCRPI Ratings and Ranks:  In the Introduction above, I referred to the various reports the NCAA publishes during the course of the season.  The Summary Report shows, for each team, its conference's NCAA adjusted RPI rating and rank.  In calcuating this number, the NCAA does not apply the basic RPI formula  to a conference's Elements 1, 2, and 3.  Rather, for the unadjusted conference RPI, it uses the average of the conference teams' unadjusted RPIs.  And for the adjusted conference RPI, it uses the average of the conference teams' adjusted RPIs.  I have determined this by calculating conference average adjusted RPI ratings myself and comparing my numbers to the NCAA's.  They match.  Given the above-described problem with the NCAA's NCRPI computations, I cannot do a similar comparison for the NCRPI.  Nevertheless, it seems reasonable to assume that for conference NCRPIs, the NCAA computation method is the same as for conference RPIs.  In other words, a conference's unadjusted NCRPI is the average of its teams' unadjusted NCRPIs.  And, a conference's adjusted NCRPI is the average of its teams' adjusted NCRPIs.

THE NCAA'S REASON FOR USING THE NCRPI

The NCAA has stated that using the RPI to evaluate football teams would not be reasonable, since football teams do not play enough games:

"Sports like softball and baseball actually play the most games and it could be argued that [their RPI ratings] are the most accurate because the sample is larger.  Soccer falls somewhere in the middle of the RPI sports in terms of the number of games.  A football RPI would be very difficult to use since each game would have an enormous impact on a team's rating."  Frequently Asked Questions About the Women's Soccer Rating Percentage Index, by Jim Wright, NCAA Director of Statistics and one of the creators of the RPI, and Rick Campbell, NCAA Assistant Director of Statistics.  The NCAA published the FAQ many years ago.

Since soccer teams' 8 to 9 non-conference games are fewer than the number of games football teams play, the NCAA's conclusion that football teams do not play enough games to make use of the RPI reasonable for football suggests that using the NCRPI to evaluate individual soccer teams might not be reasonable.  So what is the purpose of the NCRPI?

A January 23, 2009 Memorandum from the NCAA's Associate Director of Statistics to the Division I Men's Basketball Committee states the NCAA's rationale for using the NCRPI.  This is the only publicly stated rationale I've been able to find.  The Memorandum states:

"Teams and conferences are ranked by two methods.  One includes all Division I games, the other is based on non-conference Division I competition only.  Since conferences do not have the same number of members, a non-conference rating better evaluates the strength of each conference.  For example, if a conference has 10 members and plays a [double] round robin schedule, then 18 of each of its teams' 28 games (180 total conference games) will produce virtually the same won-lost percentages in Factors II and III.  But a conference with six members will produce the same numbers for only 10 of each of its teams' 28 games (60 total conference games)."  (Emphasis added.)

This rationale for using the NCRPI to evaluate conference strength is based on the way the RPI calculates strength of schedule ("Factors II and III" in the above quote).  How does this relate to Division I women's soccer?  The following two tables address this question.

This table uses data from 2013 through 2024 (excluding Covid-affected 2020).  It starts with 2013 because that is the year of the most recent major realignment of conference memberships (prior to 2024).  The yellow highlighted column shows the average percentage of in-conference games for each conference.  As you can see, there is a pretty large disparity among the conferences, from 62.4% for the Big Ten to 44.4% for the Ivy League.

These percentages are significant in relation to how the RPI formula calculates strength of schedule.  Consider Teams A, B, and C from a conference that plays a full round robin of conference games.  Teams B and C play each other, with Team B winning.  When Team A plays Team B, Team B will contribute its winning percentage to Team A's Opponents' Winning Percentage.  This winning percentage will include Team B's win over Team C.  Similarly, when Team A plays Team C, Team C will contribute its winning percentage to Team A's Opponents' Winning Percentage.  This winning percentage will include Team C's loss to Team B.  Thus the net effect of Team A's games against Teams B and C will be to pull Team A's Opponents' Winning Percentage towards 0.500 (Team B's win over Team C balanced by Team C's loss to Team B).  This pull towards 0.500 will occur for all conference games, for all conference teams.

In addition, this "pull towards 0.500" effect will carry over to Team A's Opponents' Opponents' Winning Percentage, since each of its conference opponents will have multiple other conference opponents.

As discussed on the RPI: Formula page, the effective weights of the RPI elements are Winning Percentage 50% and Strength of Schedule 50%, with Strength of Schedule comprised of Opponents' Winning Percentage 40% and Opponents' Opponents' Winning Percentage 10%.  Thus the pull from conference games of Opponents' Winning Percentage and Opponents' Opponents' Winning Percentage towards 0.500 has a variable effect on half of what determines teams' RPI ratings, with the variation depending on the percentage of conference games teams play.  The following table, again based on data from 2013 through 2024, shows the combined effective weight of the Strength of Schedule elements that pull ratings towards 0.500 for each of the conferences.

To paraphrase the NCAA's rationale for the NCRPI, its purpose is to eliminate the effects of these differences in pulls towards 0.500.

EVALUATING THE NON CONFERENCE RPI'S PERFORMANCE

According to the NCAA's above-stated rationale for using the NCRPI -- to evaluate conference performance -- the conference RPI ratings and ranks referred to in the preceding paragraph are not the best way to evaluate conference performance.  Rather according to the NCAA's rationale, it is better to evaluate conference performance using the conference NCRPI.   But is this true?  Indeed, is the NCRPI useful, at all, for Division I women's soccer?

Using 2024 as an example, the following table shows the difference between conferences' 2024 NCAA adjusted RPIs and adjusted NCRPIs as measures of conference strength:

As the table shows, for many conferences there is little difference between its NCAA RPI and NCRPI ranks.  On the other hand for the Mid-American, there is a drop from #6 using the NCAA RPI to #11 using the NCAA NCRPI.  (Note: The NCAA's conference NCRPI ranks, based on flawed calculations as discussed above, have the Mid-American ranked as #17.)  This is a significant drop.  The MidAmerican did not get any at large selections for the NCAA Tournament, but this probably is due to its not having any teams within the NCAA RPI Top 57 other than its conference champion Western Michigan -- the RPI Top 57 is the group of teams from which all at large selections historically have come.  On the other hand, Western Michigan was well within the range of NCAA Tournament seed candidates, but did not get a seed.  If the Committee considered the MidAmerican's NCRPI rank, it could have played a role in that decision.

As stated above, however, the question is whether the NCRPI really is better than the RPI for evaluating conference performance, indeed should the Committee use the NCRPI at all?  To answer this question, I use a 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 the NCRPI as compared to the RPI.  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 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 and calculations underlying the test.

ALL TEAMS AS A GROUP

As this table shows, the NCRPI's consistency of ratings with actual results is poorer than the RPI's.  From a rating system perspective, the difference between how the RPI and NCRPI perform is very large.  In my opinion, this difference is large enough to disqualify use of the NCRPI to evaluate individual teams.

TOP 60 TEAMS

Again, the NCRPI's consistency of ratings with actual results is poorer than the RPI's.  And again, from a rating system perspective, the difference between how the RPI and NCRPI perform is very large.  Here too, in my opinion the difference is large enough to disqualify use of the NCRPI to evaluate Top 60 teams.

INDIVIDUAL TEAMS, ACTUAL LESS LIKELY WINNING PERCENTAGE DIFFERENCES

As you can see, for individual teams the NCAA NCRPI does a much poorer job than the RPI of fairly rating teams in relation to each other.  The differences between teams' actual results and their likely results based on their ratings are much greater both for individual teams and for all teams as a group.  Again, in my opinion the difference is large enough to disqualify use of the NCRPI to evaluate individual teams.

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

As this table shows, the NCRPI has a slightly bigger problem than the RPI in terms of the difference between how it ranks teams and how its formula ranks teams as strength of schedule contributors.  This means it is slightly easier for smart scheduling to "trick" the NCRPI.

CONFERENCES, OVERALL PERFORMANCE FOR CONFERENCES

As you can see, the Spread is poorer for the NCRPI than for the RPI, but the Over and Under Total is better for the NCRPI.

CONFERENCES, RELATIONSHIP BETWEEN CONFERENCE PERFORMANCE AND CONFERENCE RATING

The following chart is for the NCAA RPI:

 Compare this to the comparable chart for the NCAA NCRPI:

The two charts are quite similar, but the trend line for the NCRPI is not quite as steep as for the RPI.  In addition,  the R-squared value for the NCRPI is a little higher than for the RPI. 

The following table will help with a comparison of the two charts:

As you can see, both the RPI and the NCRPI discriminate against stronger conferences and in favor of weaker ones, but the NCRPI discriminates a little less than the RPI.

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

The following chart is for the RPI:

Compare this to the chart for the NCAA NCRPI:

The trend line for the NCRPI suggests far less discrimination by the NCRPI than the RPI against conferences with high levels of parity.  In addition, the NCRPI's R-squared value of roughly .03 indicates almost no relationship between conference performance and in-conference parity. 

Since the NCRPI eliminates in-conference parity from the calculations, the two charts and table -- including the trend lines and R-squared values -- appear to confirm that the RPI indeed discriminates against conferences with high in-conference parity and in favor of conferences with low parity.  The NCRPI, by taking in-conference parity out of the equation, eliminates this discrimination.

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

The following chart is for the RPI:

Compare this to the chart for the NCAA NCRPI:

A comparison of the two charts indicates more discrimination by the NCRPI than by the RPI.  The NCRPI has a lesser R-squared value than for the RPI, but still a relatively high value.

This table shows that the discrimination caused by the differences, between conferences' RPI and NCRPI ratings and their ratings as strength of schedule contributors, are greater for the NCRPI than for the RPI.  The relative high R-squared values for the charts' trend lines reinforces the conclusion on the RPI: Measuring the Correlation ... page that the main driver of discrimination is how the RPI and NCRPI compute teams' ratings as strength of schedule contributors.

REGIONS, OVERALL PERFORMANCE FOR REGI0NS

As you can see, for regions the Spread is very slightly poorer for the NCRPI and the Over and Under Total is poorer.

REGIONS, RELATIONSHIP BETWEEN REGION PERFORMANCE AND REGION RATING

The following chart is for the NCAA RPI:

 Compare this to the comparable chart for the NCAA NCRPI:

You can see that the trend line is steeper for the NCRPI than for the RPI, meaning the NCRPI shows more discrimination among regions in relation to region strength.  In addition, the R-squared value for the NCRPI is higher, meaning the relationship between regions strength and discrimination is stronger for the NCRPI than for the RPI.

As you can see, both the RPI and the NCRPI discriminate against stronger regions and in favor of weaker ones, but the NCRPI discriminates a fair amount more than the RPI.

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

The following chart is for the RPI:

 Compare this to the comparable chart for the NCAA NCRPI:

The trend line for the NCRPI suggests a little less discrimination by the NCRPI than the RPI against regions with high levels of parity, but both show discrimination.  The NCRPI's R-squared value of roughly 0.44 as compared to the RPI's 0.66 indicates a weaker relationship for the NCRPi than for the RPI between region performance and in-region parity. 

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

The following chart is for the RPI:

 Compare this to the comparable chart for the NCAA NCRPI:

A comparison of the two charts indicates a lilttle lessdiscrimination by the NCRPI than by the RPI.  The RPI has a high R-squared value of roughly 0.93.  The NCRPI R-squared value is less, but still is a relatively high 0.75.

FINAL THOUGHTS ON THE NON-CONFERENCE RPI

Here are my thoughts related to the NCAA NCRPI:

1.  The computation errors that have found their way into the NCAA's calculations of the end-of-regular-season NCRPI are a clear and significant problem that must be fixed.

2.   The above comparisons suggest that the NCRPI, if better than the RPI at measuring conference performance, is barely better.  They also indicate that under no circumstance should the Committee use the NCRPI to evaluate individual teams' performance.

3.   Although the NCAA has indicated, as described above, that the NCRPI is a better measure of conference performance than the RPI, the Summary Report the NCAA provides to the Women's Soccer Committee, for use in the NCAA Tournament seeding and at large selection process, shows the individual teams' conference RPI ratings and ranks rather than the teams' conference NCRPI ratings and ranks.  This encourages the Committee to refer to conference RPIs when looking at conference strength rather than at conference NCRPIs.   If the NCRPI indeed is the better measure of conference strength, then the Report should show the conference NCRPI ratings and ranks.

4. The Summary Report shows the individual teams' NCRPI rankings.  Since the NCRPI is not useful for evaluating individual teams' performance, this information should not be in that Report.  In addition, it should not be in the Non-Conference Rankings Report.

5. As a generalization, Committee members should understand that the NCRPI has no value in evaluating individual team performance but that the NCAA considers the conference NCRPI as a better measure of conference performance than the conference RPI.

6. Apart from these thoughts, it is questionable whether the NCRPI is useful for any purpose.  The Committee's seed and at large selection process would lose little, if anything, if the Committee were to stop using the NCRPI altogether.