RPI: Strength of Schedule Problem

[NOTE:  In 2022, a new NCAA rule eliminated overtimes except in conference tournament and NCAA tournament games.  Unless stated otherwise, information on this page is based on data from 2010 to the present, with game results adjusted as though the 2022 change to no overtimes had been in effect.  These game results adjustments are possible because the NCAA data system since 2010 has shown which games were decided in overtime.  Also, the Covid-affected 2020 season is excluded from the data.]  

Updated April 2024

The purpose of this page is to give information on a problem that the RPI strength of schedule formula causes.

As described in detail on the "RPI: Formula" page, the unadjusted RPI has two main components.  The first component is a team's Winning Percentage.  The second component is a team's Strength of Schedule, determined from the formula (2 x Average of Opponents' Winning Percentages) + (Average of Opponents' Opponents' Winning Percentages).  The overall RPI formula puts these components together to produce the unadjusted RPI:

URPI = (Winning Percentage + (2 x Average of Opponents' Winning Percentages) + Average of Opponents' Opponents' Winning Percentages)/4

As explained on the "RPI: Formula" page, the effective weight ratio resulting from this formula is approximately 5: 4: 1.  What this means is that for Opponent A, its RPI rating and its contribution to another team's Strength of Schedule are as follows in terms of effective weight:

RPI Rating:  50% Winning Percentage, 40% Opponents' Winning Percentages, 10% Opponents' Opponents' Winning Percentages

Strength of Schedule Contribution:  80% Winning Percentage, 20% Opponents' Winning Percentages

In an "ideal" rating formula, an opponent's rank as a strength of schedule contributor will match its rank based on its rating.  As the above discussion shows, however, the formulas for a team's rating and for its strength of schedule contribution are quite different.  The result is that teams' RPI ranks and their ranks as strength of schedule contributors do not match.

How much difference is there between a team's RPI rank and its rank as a strength of schedule contributor?  The following table gives an answer, based on data from 2013 to the present.  I am using 2013 as the beginning of the period because I will be showing information related to the average difference for each conference, and 2013 was the year of completion of the last (prior to 2024) major conference membership realignment.

This table shows that there is a large disconnect between teams' NCAA RPI ranks and their ranks as Strength of Schedule contributors.  The average difference between these ranks is 30.9 positions, the median is 23 positions, and the largest difference is 162 positions.  Further a little under two-thirds of teams have a difference greater than 15 positions and a little under  half have a difference greater than 25 positions.

Assuming that teams' RPI ranks are more accurate reflections of teams' strength than are their Strength of Schedule ranks, what this means is that the RPI formula's Strength of Schedule calculations are significantly inaccurate in most cases and highly inaccurate in many. 

Here is a table that shows the strength of schedule contributor rank problem, by conference:

The table shows, by conference, the average difference between its teams' NCAA RPI ranks and their ranks as contributors to opponents' strength of schedule.  A negative difference means that the SoS contributor ranks are poorer than the NCAA RPI ranks.  A positive difference means that the SoS contributor ranks are better.  As you can see, strong conferences' teams, on average, are poor contributors to opponents' strengths of schedule, relative to the teams' NCAA RPI ranks.  Weaker conferences' teams, on average, are good SoS contributors relative to their NCAA RPI ranks.  Further, if you compare this table to the data on the RPI: Measuring the Correlation Between Teams' Ratings and Their Performance page, you will see that the pattern in this table is similar to the NCAA RPI pattern of conference discrimination shown on that page.

The next table shows the strength of schedule contributor rank problem, by region:

As this table shows, the strength of schedule contributor rank problem has the most negative effect on the West region and the most positive effect on the North.  Further, if you compare this table to the data on the RPI: Regional Issues page, you will see that the pattern in this table is similar to the NCAA RPI pattern of regional discrimination shown there.

The next table shows, by individual team,  the average difference between the NCAA RPI rank and the rank as a strength of schedule contributor for two sets of teams.  At the top are the 50 teams with the greatest negative differences: their ranks as SoS contributors are poorer than their RPI ranks.  They mostly are middle and lower level teams in strong conferences. At the bottom are the 50 teams with the greatest positive differences: their ranks as SoS contributors are better than their RPI ranks.  Many are upper and middle level teams from weaker conferences.

A result of this problem is that teams hoping for NCAA Tournament at large selections are under pressure to try to "game" the RPI system by scheduling non-conference opponents whose SoS contribution ranks are likely to be artificially high compared to their RPI ranks.  It also can make it difficult for middle and lower level teams from the top conferences to schedule strong non-conference opponents.  They aren't good teams to play from a SoS contributor rank perspective.