July 21,
2009
Proposal for Changes in
Computing RPI  Part 2
Greg
Van Zant, the head coach at West Virginia University, recently
sent the following recommendation to the NCAA Division I
Baseball Committee to change the formula for calculating the
Ratings Percentage Index (RPI).
The Weighting of the Three RPI Factors
As you are aware, the RPI formula consists of
three factors:
1. Division I winning percentage
2. Opponents average winning percentage
3. Your Opponents, Opponents winning
percentage
Originally in 1984, the formula was 204040 and
in 1994 was tweaked to the current 255025 for factors 1, 2 and
3. One of the reasons for reducing Factor 3 from 40 to 25 was
that mathematicians confirmed that too much weight placed on
Factor 3 gave weaker teams in strong conferences an inflated RPI
while penalizing strong teams in weaker leagues for having to
play the teams in their league each season.
Many coaches and others in the baseball community
are convinced that the current RPI formula overrewards average
teams in strong conferences and excessively hurts strong teams
in weaker conferences. This is because 75% of the current RPI
is based on who you play and only 25% is based on your success.
Teams control their own success but have only limited control of
their schedule. 21 to 30 games a year are dictated by
conference affiliation and most of the nonconference opponents
are based on where a school is geographically located.
It is very unfair that some teams currently have
a built in advantage every year with the RPI due to their
conference affiliation or geographical location.
Major League Baseball uses the RPI weighting of
10000 to rank their teams even though they play unbalanced
schedules. As you know, MLB teams in weaker divisions have an
easier schedule than teams in tougher divisions. However, MLB
bases standings strictly on winning percentage and they do not
factor in strength of schedule at all.
The variance in schedule strength is greater for
college baseball teams than for MLB teams, so taking into
account strength of schedule to adjust winning percentage makes
sense for college baseball. But it doesn't make sense that
winning/success is only worth 25% of a team's ranking and who
you play is worth the other 75%, especially when you can't
control most of who you play.
All the RPI does is take a team's winloss
percentage and adjust it up or down, based on whether the team
played a strong or weak schedule. A team's winning percentage
and success should be worth at least half of their ranking. The
other half should be their schedule strength. With 50% of the
RPI based on winning percentage and 50% based on the strength of
the schedule, one could expect to see results like this.
Team A #1 winning percentage and #1 strength of
schedule = #1 ranked team
Team B #1 winning percentage and #302 strength
of schedule = #151 ranked team
Team C #302 winning percentage and #1 strength
of schedule = #151 ranked team
Team D #302 winning percentage and #302
strength of schedule = #302 ranked team
A 503515 breakdown for the RPI factors would
mean that the first 50% of a team's RPI would be their own
winning percentage, which is the most important factor for any
team and it is what they can control. The second factor, the
opponents winning percentage (35%) and the third factor, the
opponents, opponents winning percentage (15%) would be the
remaining 50% of the RPI which measures schedule strength.
Opponents, Opponents winning percentage is the least important
and the least controllable of the three factors and thus should
be the least weighted factor.
With the current 255025 formula, something as
far out there and uncontrollable as your opponents, opponents
winning percentage is equally important as your own team's
winning percentage.
It can be argued that all 12 teams in the top
conference are better than all of the other top teams from every
other conference. It can also be argued that they aren't
better. The current 255025 gives a team that finishes with a
losing record in a strong conference too much of a bump up in
the RPI. The current RPI formula can rank a team very high that
lost all year long in its own conference. This team may be
strong, but they had all year to beat the top teams in their own
league and they couldn't do it. We need to adjust this RPI
formula so that teams that have success in other conferences can
have a chance in the postseason to do what those bottom teams
in top conferences tried to do all year and couldn't.
How about Team A who plays its natural rival and
closest Division I opponent, Team R, four midweek games each
year due to academics, economics and common sense. In year one,
Team R has a winloss record of 3025 and the next year Team R
is 1540. Team R's influence on Team A's RPI is going to be
much more positive in year one than in year two.
Assuming that Team A had identical teams in years
one and two, and everything else was equal, Team A would be
ranked higher in the RPI in year one. Does the fact that Team R
had a bad year mean that Team A is a weaker team? Should the
strength of another team, which you have no control over,
determine your ranking? This is a big reason that the total
percentage of the RPI dedicated to strength of schedule should
be reduced from 75% of the RPI (255025) to 50% of the RPI
(503515).
We need to reward success more, just like the
ABCA does with the AllAmerica Team nominations and
selections. The ABCA emphasizes that the AllAmerica Teams are
to recognize success instead of pro potential. We need to
reward Division I teams more for the success they have and less
for what conference they are in, where they are located or their
perceived strength due to an overinflated RPI. Winning and
strength of schedule are both important factors when ranking
teams and they should both be considered equally.
Various simulations of the RPI using different
formulas have resulted in some significant differences in team
rankings. Some teams have moved up as many as 44 spots while
others have dropped as much as 42 spots. If those two teams
were ranked equally before a certain manipulation of the RPI,
they could be ranked 86 teams apart with a new RPI formula.
This is the difference between playing in the postseason and
staying home.
Part
1  Overview
Part
3  The Weighting of Home Wins and Road Wins
Part
4  Exempt Contests vs. NonDivision I Teams; Bonus
& Penalty System
(photo by Jimmy Jones)
