RIP, RPI: The Case Against The NCAA Tournament Selection Committee's Favorite Statistic

Next week, the NCAA Tournament selection committee will pick teams based on a combination of factors that seems to change every year. Among those factors, however, is a tried-and-true measure that seems to carry significant weight, season after season: the Ratings Percentage Index (RPI), a statistical tool used to help determine the best teams in college basketball.

This is a problem.

In theory, something like the RPI is a good idea. After all, no one can watch every game, or even every game in a single conference: lock a group of well-meaning campus hoops experts in a room, and team resumes are going to be overlooked while brand name bias creeps in.

Enter the RPI. When one committee member tries to say that Oregon State's loss to Valparaiso must be bad because, really, where the hell is Valparaiso? another member can point to the numbers showing that Valparaiso is #actually good, despite playing in a tiny town in the gray hellhole that is Northwest Indiana.

Here's how your favorite school's RPI score is calculated:

●Winning percentage (25 percent) — home wins are weighted 0.6, neutral wins 1.0, and road wins count 1.4, and road losses are 0.6, neutral losses 1.0, and home losses 1.4.

●Opponents' winning percentage (50 percent)

●Opponents' opponents' winning percentage (25 percent)

That's it. That's the formula—and it isn't good enough. Sure, wins and losses are pretty important. Strength of schedule matters, too: going undefeated in the Big 12 is better than doing the same in the SWAC.

Thing is, the RPI doesn't account for how teams win and lose. And it overvalues schedule strength, too. Taken together, these overlapping factors create unnecessary and unfair distortions, both when it comes to schedule making and NCAA Tournament seeding.

Want to look good to the NCAA selection committee come March? Well, the RPI doesn't take into account games against non-Division I teams, and since opponent winning percentage is so overrepresented, it's often better to play games you know you'll probably lose rather than games against more evenly-matched opponents. Moreover, don't take games against teams you're clearly better than.

Here's an ideal scheduling philosophy:

●Instead of racking up easy wins against the Grambling States of the world, play a Division II school nobody has ever heard of. The Grambling game counts against you. The Division II game does not.

●Schedule games against top opponents. Lose? It doesn't matter. Win? Your RPI goes up.

●Schedule the best teams in bad conferences. NCAA Tournament bubble teams shouldn't have a problem beating the best team in, say, the NEC, so that game shouldn't count much more than beating the worst team in the NEC. But it does, because their winning percentages will be so different.

Okay. It's possible to defend all of the above by arguing that the RPI encourages teams to schedule tougher opponents, theoretically creating better regular season television. Fine. What's far less debatable is that the RPI misguidedly fails to account for margin of victory, which is arguably the best statistic—by far!—for determining the country's best teams.

Given that Division I features 351 teams, schedules are going to vary dramatically. So it's just not enough to just consider who teams beat. What also matters is how they beat their opponents. All wins are not created equal.

Take Wichita State. The Shockers play in the Missouri Valley Conference, which is a fine mid-major conference, but according to KenPom.com, it still ranks No. 12 nationally. That's not great.

The RPI sees that Wichita State won the conference by four games, going 16-2 in conference play. However, all the RPI knows is that the Shockers beat a bunch of inferior teams. That's not super-helpful. Were a lot of those games close, indicating that Wichita State might have gotten lucky? Were they blowouts, indicating that the Shockers took care of their schedule just as well as a top major conference team might have?

RPI punts on that question and just assumes WSU is the best among bad teams. However, ratings that take into account margin of victory show that Wichita State dominated its in-conference opponents in a manner equivalent to how top power conference teams likely would if placed in the MVC.

The result? RPI ranks Wichita State No. 48. The KenPom statistical rankings, which take into account margin of victory, place the Shockers No. 11. Big, big difference, and one to consider when filling out your brackets.

On the other side is Maryland. The Terrapins are ranked No. 12 in the RPI, but they're No. 24 in KenPom. In other words, RPI thinks Maryland should be a No. 3 NCAA Tournament seed, while KenPom thinks that would be vastly overrating the Terrapins.

The reason Maryland is so much higher on RPI is that it's getting more credit than it deserves for close wins. Since winning and losing in close games is statistically close to a random occurrence, it makes sense that Maryland should not receive full credit for barely beating teams like Georgetown, Rider, Northwestern and Nebraska.

Last year, Maryland fans got angry that KenPom ranked the No. 4 NCAA Tournament-seeded Terps so low at No. 32. As it turns out, Maryland's luck in close games was just that, and the Terps lost in the second round of March Madness after barely escaping in the first.

The NCAA argues that it excludes margin of victory because it wants to reward resumes, not the best statistical teams. That's fair, but those two things aren't mutually exclusive. Every other component of the selection process is resume-based, and even RPI isn't all about wins and losses, as its biggest component is strength of schedule, so heavily weighted that wins and losses legitimately don't matter:

— We take the stairs (@NoEscalators)February 24, 2016

Using a top-notch statistical rating system to determine the field and seeding for the NCAA Tournament is a very good idea. Someday, perhaps, the committee will agree.