The RPI. If you have ever paid attention to any NCAA Tournament bubble talk, then you have at least heard of the RPI. But do you really understand its intricacies? Do you really understand how easy it can be manipulated and how crucial it is for bubble teams to engage in said manipulation? Or put another way, do you truly realize how idiotic it is that the NCAA Selection Committee continues to use the RPI as a major component of their selection process?
Well, break out the notebook (old school or new) because class is in session. Over a series of articles, each building upon the next, we will participate in a journey that illuminates the astounding foolishness that is the RPI. You’ll laugh, you’ll cry, your mind will be boggled. And at the end of the day you’ll realize the continued use of the RPI is just quintessential NCAA.
A contextual note before we embark. There will not be a lot of time spent on other, far superior ranting methodologies. There are many out there (KenPom, BPI, Sagarin, the AP Poll, rank by mascot, pulling names out of a hat, eenie-meanie-miney-mo) and our own Michael Rogner has previously extolled the virtues of several. It is universally acknowledged that a rating system aside from the RPI is superior to the RPI. The purpose of this case study is not to debate the merits of the RPI compared to other rating systems, but rather to take an in-depth look at what makes the RPI so freaking dumb.
And here. We. Go.
RPI 101 - Not All Cupcakes are Created Equal
The Basics: The RPI, which stands for Ratings Percentage Index, is a rating system heavily used by the NCAA Tournament selection committee as a tool to select at-large teams and seed all 68 teams. Beginning around February, countless college basketball conversations will center around “RPI top 50 wins,” “RPI top 100 wins,” “Sub-RPI top 150 losses,” OOC strength of schedule (as determined by the RPI), overall strength of schedule, and a team’s own RPI rank.
The stated purpose of the RPI is to rate teams’ relative strength by using a combination of win/loss record and strength of schedule. In theory, that sounds nice. It’s impossible for every committee member to watch every single college basketball game, and having a metric that allows for team-by-team comparisons makes sense. Unfortunately, the RPI does not.
A team’s rating on the RPI is determined from three factors:
- That team’s winning percentage (25%)
- That team’s opponents’ winning percentage (50%)
- That team’s opponents’ opponents’ winning percentage (25%)
It does not factor in margin of victory. It does not factor in garbage time. It does not factor in pace of play—something that varies greatly in men’s college basketball. It does not factor in efficiency. It does not factor in injuries or suspensions.
However, within those three categories, there are several nuances that are easily manipulated and significantly alter the ratings themselves. These nuances will be a primary focus of our exploration.
When a Win is a Loss:
In Romeo and Juliet, William Shakespeare taught the world that a “rose by any other name would smell as sweet.” In other words, it doesn’t matter what you call it, a rose is still a rose. William Shakespeare would not be fond of the RPI.
There are 351 D1 teams in men’s college basketball. The bottom 150 or so of these teams are quite bad and losing to one can crush an other-wise worthy resume. Honestly, losing to any team outside the top 100 is frowned upon, but lose to a team outside the top 200 and you’ve really pooped your pants. The good news is, the caliber of teams consistently in the running for NCAA Tournament at-large selection rarely lose to these true cupcakes. For example, Florida State has only lost two games to teams outside the top 200 in the last 12 years—at Auburn in 2011 and at Boston College in 2012 (ironic since those two teams went to the Sweet Sixteen and won the ACC, respectively).
So if NCAAT quality teams rarely lose these games, why are they such a big deal? It’s a good question, with an answer that might perplex you. Go back up to factor two of the RPI, “the teams’ opponents’ winning percentage.” This represents FIFTY percent of a team’s RPI. Yes, the win/loss records of your opponents is weighed twice as heavily as your own win/loss record. Therefore, simply playing a true bottom feeder significantly impacts a team’s RPI.
Ten years ago, this was a guessing game. It was known that playing horrific teams hurt your RPI, but there was always a debate over just how much. Now, thanks to our friends at RPIForecast.com, the debate gets answered. And the results can be shocking.
Rutgers currently sits around 82 in the RPI (the exact number can change by the hour as new results roll in), outside NCAAT at-large selection range despite sporting a 10-1 record against D1 teams. A large reason for the Scarlet Knights’ low RPI despite a very good record is an abysmal strength of schedule. One game in particular that’s weighing them down is a win over 2-9 Central Connecticut State. Rutgers won this game by 42, but in the end they may have lost.
Rutgers’ projected RPI, based on computer simulations of all games, is around 124 (because Rutgers’ record is expected to get worse as they move through Big 10 play. Current RPIs are based only on games played up to that moment in time) and their projected SOS is 92. If Rutgers dropped Central Connecticut game completely from their schedule and did not replace it, their expected RPI and SOS would move up to 115 and 80, respectively. Yes, you read that correctly. Removing a 42 point win would enhance Rutgers’ resume. Dropping Central Connecticut and adding a home loss against Gonzaga results in an RPI of 115 and a SOS of 69. Rutgers finishes with a better RPI if they replace a win with a loss.
Now, an RPI of 115 is still not going to result in an NCAAT at-large selection. But what if that nine spot increase was from 62 to 51? That could be the difference between a 12 seed in the Dance and a 2 seed in the NIT. Taking away a 42 point win and replacing it with a loss could be the difference between an NCAAT bid and a burst bubble. It could be the difference between a coach fired and a contract extension.
Cupcakes and Faux-cupcakes:
This brings us back to Mr. Shakespeare. A rose by any other name still smells sweet, but in RPI-land not all cupcakes are created equally.
Central Connecticut is a bad basketball team. Any metric will confirm this statement. Kenpom ranks them 340th out of 351 teams. But being a bad team, in and of itself, is not what causes these Blue Devils to be an RPI-anchor. Hear me out.
People often refer to RPI-anchors by their actual RPI. “FSU’s RPI is being dragged down from playing ‘xyz sub-250 team’” is a phrase I see frequently. And in practice, this is likely true. However, the reality is, the opponents’ RPI is not what drags down your own RPI. Go back to the three factors above and say it with me again: the opponents’ winning percentage is 50% of your team’s RPI. And RPI ranks do not necessarily align with winning percentages (for reasons we will discuss later in the series).
Texas Southern’s current RPI is 51. Their record is 4-9. If FSU played the Tigers tomorrow, nine losses would be added to the Seminoles’ opponents’ winning percentage ledger against only four wins, resulting in a net loss of 5 games. Meanwhile, Oakland’s current RPI is 216, but their record is 7-3. Playing the Grizzlies tomorrow would result in a net gain of 4 games.
Results like this litter the entire index. And they often have zero correlation with how good the opponent actually is. That’s because some cupcakes play in leagues with teams that are even worse than they are. Tennessee-Martin might be 221st in KenPom (meaning the odds of an average bubble team team losing to them are quite low) but they are still projected to finish with a 19-12 record, in part because they are better than many of the other teams in their league. Conversely, Oregon State ranks 19 spots higher on KenPom, but because the Beavers play in the Pac-12 they are projected to finish 7-24 on the season.
Both Oregon State and UT-Martin would likely be blown out by FSU (again, the ‘Noles have lost 2 games to sub-200 KenPom teams in the last 12 seasons), but Oregon State would be a net-loss of 17 games on the opponents’ win percentage, while UT-Martin would be a net-positive of seven. Indeed, a cupcake by any other name does not taste as sweet.
In the next installment of this case study in idiocy we will take a look at what I like to call the “Real Estate” rule.