KenPom vs. BPI: Where RPI Always Loses

Over the course of the next two weeks leading up to Selection Sunday, you are going to here about RPI over and over.

RPI is a major tool used by the selection committee, measuring team strength for purposes of at-large picks and seeding. The Rating Percentage Index ranks teams by their wins, losses, and strength of schedule, and the NCAA has used it since 1981 for March Madness. The core part of RPI’s formula is ¼ team’s winning percentage, ½ opponents’ winning %, and ¼ opponents’ opponents’ winning %.

Image

The tool ignores margin of victory altogether – and it has been challenged, especially of late, with new metrics created by men like Ken Pomeroy (KenPom) and Dean Oliver (ESPN’s BPI).  RPI relies too heavily on a team’s winning percentage, which is a problem because college basketball schedules are not created equally. In addition, it is easy for colleges to try to beat the system by playing good teams (regardless of outcome of the game), by avoiding teams bound to have bad records, and by playing at home as much as possible. If you follow those simple steps when making your team’s schedule, you will be rewarded with a strong RPI.

The RPI also unfairly rewards “power conferences” over their mid-major counterparts. Mid-Major teams are hurt by the fact that teams they play are typically weaker than those that major conferences teams play. Since margin of victory does not matter, a blowout loss at the hands of a top ten team is more beneficial than a 30-point win over a weak team. It is exceedingly more difficult for a mid-major to make a schedule that will be reflected positively in RPI than it is for a member of one of the power conferences.

That’s enough about RPI. Hopefully, you clearly see its flaws and understand the need for a new metric the tournament committee can rely on.

One such metric could be the College Basketball Power Index introduced this year by ESPN and Dean Oliver. While the RPI does not incorporate (1) scoring margin, (2) diminishing returns for blowouts (a 30-point win is not necessarily twice as good as a 15-point win), (3) pace of game, (4) strength of schedule beyond opponents’ opponents’ wins and losses, and (5) de-weighting games with missing key players, ESPN’s BPI does. In addition, BPI does everything that RPI does. It is an incredible system that reflects teams’ résumés very well. As Mr. Oliver puts it, BPI “explains a team’s wins and losses in a more detailed context and seemingly predicts future results as well as other tools, if not better. It also incorporates information about injuries and misses games that should be relevant for weighting the résumés of bubble teams on Selection Sunday”. It should be easy to see why I am such a big fan of BPI (and Dean Oliver).

Image

Another more well known metric is Ken Pomeroy’s efficiency rankings on his website. Mr. Pomeroy’s rankings, as opposed to Mr. Oliver’s, are purely predictive. His website is also incredible (and available for a yearly subscription of just $20) with seemingly endless amounts of information, but we are only going to discuss his Pythagorean calculation for expected winning percentage (originally developed by the great Bill James in baseball). Inputted into his formula are the adjusted offensive and defensive efficiencies of each college basketball team, which according to Pomeroy are “based purely on scoring per possession with no consideration of winning or losing”. On KenPom, the term ‘adjusted’ “refers to how a team would perform against average competition at a neutral site” with more weight on recent performances. Unlike RPI, KenPom does take into account scoring margin, pace of game, and strength of schedule beyond opponents’ opponents’ wins and losses – but it does not diminish returns for blowouts and de-weight games with missing key players like BPI does.

In my mind, BPI and KenPom are the two best metrics available today in terms of looking at a team’s résumé (BPI) and predicting how the team will perform in the future (KenPom).

Today, I thought it would be interesting to look at how the two metrics are currently ranking teams. First, I wanted to look into the rankings and determine which teams Mr. Oliver and Mr. Pomeroy disagree the most on:

The biggest difference in the two rankings is regarding Wagner. The Seahawks of the NEC are 17-11 overall and 11-4 in the conference, trailing only Robert Morris. Wagner has won seven straight games, led by seniors Kenneth Ortiz and Latif Rivers. Today, KenPom ranks Wagner 176th, while ESPN’s BPI ranks them 220th (a difference of 44 spots). Both KenPom (154th) and BPI (176th) agree Robert Morris is the best team in the NEC, however.

Image

Louisiana Monroe (257th & 216th), Charleston Southern (248th & 210th), Radford (253rd & 215th), and Abilene Christian (347th &310th) follow Wagner in terms of teams with the highest difference between KenPom and BPI rankings.

The highest difference among top 100 teams surrounds Vermont. The leader of the America East Conference, Vermont is 20-9 (14-1) behind the play of Clancy Rugg, Sandro Carissimo, and Brian Voelkel. The Catamounts are battle tested, as they played at Duke, Harvard, and Providence in the non-conference. Despite a 0-3 record in those games, Vermont took Duke to the wire at Cameron Indoor, falling 91-90 on November 24. John Becker’s squad has won 16 of their last 17 – and looks ready to bring a scare to a top-seed in the tournament. KenPom currently ranks Vermont 56th, while BPI is less optimistic with a ranking of 88th.

Toledo is a team BPI has in its top 100 (81st), while KenPom has it ranked 109th. Toledo currently trails Western Michigan in the West Division of the MAC, while Buffalo is the leader in the East Division. Buffalo (97th) and Ohio (108th) are both higher rated on KenPom than Toledo. BPI, however, has Toledo as the highest ranked MAC team, better than both Buffalo (99th) and Ohio (100th).

Georgia State (71st & 95th), Boise State (83rd & 60th), and Missouri (61st & 39th) are some other top 100 teams KenPom and BPI rank very differently.

Finally, let’s look at some of this season’s better-known teams:

UMass was ranked as high as 13th in the AP Poll this season. Senior Chaz Williams picked up some All-American buzz, and the Minutemen looked primed to run through the Atlantic 10 – but they currently sit behind Saint Louis, St. Joseph’s, and VCU in the conference standings. KenPom currently has UMass ranked 44th, but BPI puts them much higher at 29th overall. UMass trails VCU (22nd), Saint Louis (25th), and George Washington (40th) on KenPom, but only VCU (21st) and Saint Louis (25th) in BPI.

There are teams KenPom thinks better of than BPI, like Tennessee (26th vs. 40th), Michigan (14th vs. 23rd), and Virginia (4th vs. 12th).

And there are teams BPI likes more than KenPom, like Kansas (3rd vs. 8th), Iowa State (13th vs. 23rd), and Kentucky (8th vs. 18th).

The list goes on and on – as does the competition between KenPom and BPI.

As we move into the future of college basketball, it is clear the selection committee should be using RPI less – but what will replace it?

A case can certainly be made for both BPI and KenPom. At Big East & Beyond, we will continue to keep tabs on how the two metrics perform against each other.

Which will do better in terms of selecting the tournament field? Which will do better in terms of picking winners within the bracket?

In the coming weeks, Zurg will look at both of these.

Did you like this post? Send us your comments below or on Twitter (@BEandBeyond)

Advertisements

Tell us what you think! Comment here!

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s