Tuesday, September 30, 2025

2025 ARTICLE 20: RPI REPORTS AFTER WEEK 7 GAMES

Before showing the regular tables in these weekly reports, I will provide some details that should be sounding alarm bells for the Women's Soccer Committee.

What Is Needed for the NCAA RPI to Work Properly as a Rating System for Teams Across the USA?

In order for the NCAA RPI to properly rate teams across the country in a single rating system, there are a number of conditions that the "facts on the ground" must meet:

1.A.  The different regions of the country, within which teams play most of their games, must be of equal strength; or

1.B.  There must be a lot of out-of-region competition.  If teams played all their games only in their own geographic region, then each region's teams' ratings would be the same as each other region's teams' ratings, for practical purposes.  This would be true regardless of whether the regions were equal in strength.  Thus if the regions are of unequal strength, the NCAA RPI depends on there being enough out-of-region competition to get each region's teams properly rated in relation to teams from other regions.

2.  Nationally, teams' ratings are distributed in a bell curve fashion.  At one end of the bell curve are relatively few teams with very high ratings and at the other end are relatively few with very low ratings. As ratings approach the middle of the rating spectrum, there are more and more teams, with the most teams at the center of the bell curve. 

Because of the structure of the NCAA RPI formula, for it to properly rate teams across the country in a single system, each region must have a similar bell curve distribution, in other words similar proportions of teams at the ends and in the middle areas of the curve.  Put more simply, the levels of parity must be the same for the different regions.  If the regions have different levels of parity, then the structure of the NCAA RPI formula will cause it to underrate the more highly rated teams from higher-parity regions and overrate the more highly rated teams from lower-parity regions.

The Women's Soccer Committee made this even more of an issue in 2024 when it changed the NCAA RPI formula so that the formula's Winning Percentage element values ties as only 1/3 of a win rather than the previous 1/2 of a win.  The proportion of in-region ties in a region is a function of the level of parity in the region, in other words of the shape of the region's rating distribution bell curve.  If each region has the same level of parity -- which would be expected to result in the same proportion of ties -- as each other region, then the change to 1/3 will effect the regions equally.  If the regions have different levels of parity, however, then the change will increase the NCAA RPI's discrimination against regions with a high level of parity by having a negative impact on those teams, since they will have more ties.

Out-of-Region Competition

With few exceptions, teams play most of their games within the geographic regions in which they are located.  I have identified four regional groups of states.  The schools in the states in each region, as a group, play the majority or plurality of their games in their region.  You can find the states in each region as well as a map showing the regions at the RPI: Regional Issues page of the RPI for Division I Women's Soccer website.

The following table shows the percentages of games teams from each region played against teams from their own regions as well as against teams from the other regions for the period 2013 to 2024:




This year, there is a big change.  The following table shows the actual regional distribution of games played so far:


As you can see from comparing the two tables, each region's teams are playing a significantly higher percentage of games in region as compared to what they have played historically.

And, when I add in the games on teams' schedules but not yet played, I get the following table:


When you compare this table to the first one above, you can see that this year there is going to be a big decline in the percentage of out-of-region games.  It will be an 18.7% decline for teams from the Middle, 28.2% decline for teams from the North, 31.1% decline for teams from the South, and 34.9% decline for teams from the West.

As I wrote above, in order for the NCAA RPI to function properly as a tool for rating teams across the nation, it must have enough cross-region games to get each region's teams properly rated in relation to teams from the other regions.  I will express an opinion here:  There will not be enough cross-region games this year to do that.

Is the Women's Soccer Committee aware of this problem?  Possibly.  If so, has it figured out how to deal with it?  I am doubtful.

Differences in In-Region Parity

As I also wrote above, the proportion of a region's in-region ties is a function of the level of parity within the region.

The following table shows the actual proportions of in-region ties so far this season, for the four regions:



And, the next table shows my predicted proportions of in-region ties when the entire season is completed.  It is likely the table slightly understates what the final proportions will be:


What these tables indicate is that when the final numbers are in, the South region likely will have a significantly lower proportion of in-region ties than the other regions,  The North seems likely to have the highest proportion, followed by the West and then the Middle.  In other words, the regions do not have the same levels of parity.  As a result, the NCAA RPI's ability to properly rate teams across the nation, already impaired by the decrease in out-of-region competition, will be further impaired due to differences in in-region parity -- an impairment even further exacerbated by the Committee's 2024 tie-value change.

Is the Women's Soccer Committee aware of this problem?  Probably not,  If so, has it figured out how to deal with it?  Probably not.

Consequences of the NCAA RPI Formula's Medthod of Computing Strength of Schedule

As I've written previously, because of the way the NCAA RPI formula is constructed, the ranks the formula assigns to teams is different than what teams' ranks are within the part of the formula that assigns them values as strength of schedule contributors to their opponents.

So that you can see how this causes a problem, I will use a current example that predicts how two teams will end up based on the actual results of all teams' games played so far and the predicted results of all teams' games not yet played:


Based on past history, teams in the Top 57 of the NCAA RPI are possible at large selections for the NCAA Tournament.  Teams ranked #58 and poorer never have gotten an at large selection.  Thus in this example, Lipscomb is a potential at large team and California is not.

As you can see, Lipscomb's opponents' NCAA RPI ranks and their ranks under the NCAA RPI formula as strength of schedule contributors are only four positions apart.  In other words, if the NCAA RPI ranks of Lipscomb's opponents are correct, then the NCAA RPI also is correctly measuring those opponents' contributions to Lipscomb's NCAA RPI strength of schedule and thus to its NCAA RPI rating.

On the other hand, that is not the case for Cal.  If the NCAA RPI ranks of Cal's opponents are correct, then the NCAA RPI is grossly understating (by 38 rank positions) those opponents' contributions to Cal's NCAA RPI Strength of Schedule and thus to its NCAA RPI rating.  In other words, the NCAA RPI is grossly underrating and underranking Cal.  It should be a candidate for an at large selection, but its NCAA RPI rank says it won't be.

How will the Women's Soccer Committee overcome this flaw in the NCAA RPI formula?  The most likely answer is, It won't.  The data reports the NCAA staff gives the Committee for use in the at large selection process contain a lot of data.  They do not, however, tell the Committee teams' ranks as strength of schedule contributors to their opponents.  Thus the Committee never is able to make the kind of comparison the above table allows.

Should the Committee ask the NCAA staff for SoS contributor rank information?  Yes.  Would the NCAA staff be likely to provide it, if asked?  I doubt it.  Is the Committee likely to find a satisfactory way to deal with this problem?  Probably not.

THIS WEEK'S TABLES

Below are the following reports, after completion of Week 7 of the season:

1.  Actual Current Ranks.  These are RPI reports based only on games already played.  Teams' actual ranks in these reports (and the ratings on which the ranks are based) exactly match those published by the NCAA at the NCAA's RPI Archive, and also those published at Chris Henderson's 2025 Division I College Women's Soccer Schedule website.  These reports also include teams' current KPIMassey, and Balanced RPI ranks so you can see how the different rating systems compare.

2.  "Predicted" End-of-Season Ranks.  These are RPI reports based on the actual results of games already played PLUS predicted results of games not yet played.  The purpose of these reports is to give an idea of where teams might end up at the end of the regular season. The reports show both NCAA RPI and Balanced RPI ranks.

The result predictions for future games use teams' actual current NCAA RPI ratings as the basis for the predictions.  So these reports show where teams will end up if they all perform exactly in accord with their current NCAA RPI ratings.  As each week passes, the predictions come closer and closer to where teams will end up.

ACTUAL CURRENT RANKS

 Here are the actual current NCAA RPI and Balanced RPI ranks for teams.  For an Excel workbook containing these data, use the following link: 2025 RPI Report Actual Results Only After Week 7.

NOTE:  If you use the link, you will see the workbook in a Google Sheets format, which will be difficult or impossible to read.  Rather than trying to use that workbook, take the following steps to download the workbook as an Excel workbook:

Click on File in the upper left.

In the drop down menu, click on Download.

In the drop down menu, click on Microsoft Excel (.xlsx).

This will download the workbook as an Excel workbook.

In the tables, be sure to note the differences between teams', conferences', and regions' NCAA RPI ranks and their ranks, within the NCAA RPI formula, as strength of schedule contributors to their opponents' ratings.  You also can see the same information for the Balanced RPI.

Also, for each of teams, conferences, and regions, these reports show current KPI and Massey ranks so you can compare them to the NCAA RPI and Balanced RPI ranks.

In the Teams table, the color coded columns on the left show, based on past history, the teams that are potential seeds and at large selections for the NCAA Tournament, given their NCAA RPI ranks at this point in the season.


Here are the actual current ranks for conferences:


And here are the current actual ranks for the regions.


"PREDICTED" END-OF-SEASON RANKS

Here are the predicted end-of-season NCAA RPI  and Balanced RPI ranks for teams.  For an Excel workbook containing these data, use the following link: 2025 RPI Report After Week 7.

The color coded columns on the left show, based on past history, the teams that would be candidates for NCAA Tournament seed pods and at large positions if these were the final NCAA RPI ranks.


Here are the predicted end-of-season ranks for conferences:


And here are the predicted end-of-season ranks for the four geographic regions:



No comments:

Post a Comment