More Than Meets the RPI: The RDW (RPI Doesn’t Work) Invitational | Stat Geek Idol

March 19, 2012 - by Bill Bihn

***NOTE*** This is a Round of 64 entry in our inaugural Stat Geek Idol contest. This article was conceived of and written by Bill Bihn. The opinions or predictions expressed below do not represent the views of TeamRankings.com, and are solely those of the author. 

In the beginning of March, experts from around the country begin to analyze who they feel belongs in the NCAA Basketball Tournament. Every time you turn on the radio or television, one gets beaten over the head with statistics that supposedly prove which “bubble” team deserves to be included in the field of 68. Of all the numbers that get thrown around, the most common one almost certainly has to be RPI (Rating Percentage Index).

“No team with an RPI over 80 has ever earned an at-large bid.”

“No team with an RPI under 20 has ever failed to get into the NCAA tournament.”

“If your RPI is between 46 and 49 and your coach’s first name includes the letter Y, there is a 37% chance you will be selected as an 11-seed.”

Most non-statisticians who hear RPI numbers are perfectly accepting of the number. After all, the four 1-seeds in the 2012 NCAA tournament are also the teams with the top four RPIs. What could be wrong with the RPI? It turns out the RPI has a dark secret—a secret most novice sports fans have yet to discover.

The RPI is severely flawed.

Do you need proof? Let’s take a quick look, Vegas style.

According to Team Rankings, as of Selection Sunday Southern Mississippi had an overall RPI rating of 22nd. Kansas State had an overall RPI of 46th. The RPI formula deemed Southern Mississippi to be the significantly better team. And yet, Kansas State was a 5.5-point favorite by the time their 2012 tournament matchup tipped off. If sportsbooks (the ones putting their money where their mouths are) give no credence to RPI numbers, chances are high that you shouldn’t either.

The “RPI Doesn’t Work” Invitational

To get a closer analysis of why the RPI doesn’t work, let’s create a pretend postseason tournament called the 2012 “RDW (RPI Doesn’t Work) Invitational.” In this invitational, we will focus on two teams, which we will “creatively” call Team A and Team B. Both of these teams will play five games. Team A’s five games will be home matchups against Kennesaw State, UNLV, and Penn State, and road games against Michigan State and Ohio State. Meanwhile, Team B will host New Mexico State, UAB, and Central Florida and travel to Arizona State and Memphis.

Let’s say both teams went 4-1 in this 5-game tournament. Team A’s only loss was to Michigan State, and Team B lost to Memphis. Overall, both teams went 3-0 at home and split their two road games. Which team had the better performance?

Clearly, the nod should be given to Team A. They played two top-10 teams, Michigan State and Ohio State, on the road, and beat one of them! In addition, they also had a quality top-25 victory at home against UNLV. Taking care of business against two lowly teams, Penn State and Kennesaw State, isn’t much, but they did win those games. Team B, meanwhile, only played one top-50 team (Memphis, who isn’t ranked in either national poll). They lost that game.

Even though Team A was by far the better team in the hypothetical tournament, Team B is better according to the RPI formula. That’s right—apparently playing two top-10 teams and a third from the top-25 isn’t good enough to top a schedule loaded with quality teams such as UAB (RPI: 110) and Arizona State (RPI: 252).

In case you haven’t noticed by now, the “RDW Invitational” did actually happen. Five example games were taken from the 2011-2012 schedules of Wisconsin (Team A) and Southern Mississippi (Team B). And just as in the example, the RPI claims Southern Mississippi is a better overall team, top to bottom, than Wisconsin this year. The 5-game capsule described above is an example of why.

The RPI Punishes Blowouts Over Weak Opponents

If you are curious as to how this statistical travesty actually occurred, it all came back to Kennesaw State. The 2-28 Kennesaw State Owls had such a low winning percentage this season that even beating them sunk Wisconsin’s RPI substantially. And since the RPI does not have a “margin of victory” component, there was no way to remedy this (not even an 85-31 blowout, which is what actually took place).

The ramifications of using RPI in college basketball could be substantial. Since even a win against Kennesaw State causes a team like Wisconsin to lose points in the RPI, one would think the only option is to not schedule a team like that at all in the future. However, this has a huge effect on the economics of college athletics. Kennesaw State keeps its program afloat because of paydays like the one they received from going to Wisconsin. Games like this are a win-win; Wisconsin gets to sell 17,230 tickets [PDF] and Kennesaw State receives a huge monetary “thank you.” If Wisconsin and other large programs stopped scheduling these games simply for fear of an RPI meltdown, the entire landscape of small-school Division I basketball could change.

How is it that a statistic highly regarded among circles of experts (not to mention the men and women in the tournament selection room) can be so misguided? How could it be that RPI is shoved down our throats, despite its huge flaws?

Only You Can Prevent RPI Abuse

What should we as fans, experts, and bracketologists do about this?

In the end, we must not allow ourselves to fall into the trap of believing one story. We must check multiple sources and come to our own conclusions. We must stop believing Southern Mississippi is the 22nd best team in the country if Jeff Sagarin thinks they are 57th best and Ken Pomeroy has them ranked 71st. We must make ourselves aware of Colorado State’s flaws (RPI: 29, Sagarin: 76, Pomeroy: 76) and likewise give credit where credit is due, such as in the case of Miami of Florida (RPI: 66, Sagarin: 44, Pomeroy: 38).

There’s more to life than meets the eye, and there’s more to college basketball than the RPI.