Replacing the RPI

Sagarin's formulas most credible, head to head

My weeks-long exploration of the numerological netherworld of the Ratings Percentage Index, which appears in this week's SI, was a journey into terra incognita.

Like Brooke Shields, I somehow got through Princeton without taking a math course. We have a routine in the Wolff household: Whenever a task involves arithmetic, my wife insists that I do the calculations twice. Yet here I was, sorting through decimal points, and terms like "Bayesean" and "retrodictive," and computer printouts long enough to make a geek swoon.

I'm only marginally more math literate now than when I took the assignment. But I understand the RPI much better, and my hunch going in-that college basketball would be better off without it-has hardened into a conviction.

Bracketologist Joe Lunardi, on espn.com, recently gave people like me the green light. "I have no problem with those who criticize the RPI," he wrote, acknowledging that the metric is full of flaws. (He didn't enumerate them, but I'll quickly tick off three: The RPI ignores margin of victory; it ignores the home edge; and it ignores common sense by sometimes sending a team that wins down and a team that loses up.)

I have enormous respect for Dr. Joe's basketball instincts and the man-hours he devotes to his craft. Yet he then launched into a defense of the RPI, culminating in this: "For those who suggest doing away with the RPI, I ask, 'To be replaced with what?'"

The answer is in this week's SI, and I'll recapitulate it here. The RPI should be replaced by any of the three metrics developed by Jeff Sagarin, whose rankings can be found at http://www.usatoday.com/sports/sagarin.htm.

One, based on the chess-rating system of Hungarian émigré physicist Arpad Elo, confines itself to wins and losses. A second is Sagarin's Predictor, which is based on simple point differential. In his third, the composite Rating, Sagarin weights Elo and the Predictor equally.

Dr. Joe believes that Sagarin's Rating puts too much stock in point differential. Fine. He can always use Elo, which ignores how much teams win and lose by. Elo is a kind of Al Davis metric -- just win, baby -- yet at the same time it (like all of Sagarin's ratings) takes into account the twin elephants in Division I basketball's living room: that the home team wins two of every three games it plays; and that teams from the power conferences refuse to play at the little guys' places.

Of course, plenty of fans actually like a rating system that takes into account points for and against, and for them Sagarin offers the Predictor. Sure, you can find oddities early in the season, when a blowout or two might skew the results. But by Selection Sunday, Sagarin's sample is large enough that the Predictor has an accrued credibility. More credibility, I believe, than even Elo and the Rating. And much more than the RPI.

How do we know? Because when we have nearly three decades to go by, it's easy to determine the superior system. It's the metric that best predicts what will happen when two teams play each other. We're not talking theory here-no gentlemen in private clubs, drumming their fingers on mahogany tables and stroking their chins while Jeeves pours the cognac. We're talking results: Over the 28 years for which Sagarin has assembled data for the RPI and his three metrics, the RPI fared worst of the four in forecasting how teams actually did in postseason play.

Sagarin's adaptation of Elo-a system vetted by the mathematical minds that populate the world of chess-beat the RPI in 22 of those 28 seasons. Sagarin's composite Rating beat it 24 times. And Sagarin's pure-points Predictor went 26-2.

End of discussion.

Sure, if the bracketologist's primary job is to suss out what the draw will look like, the RPI has to be a tool in his toolbox. After all, it's a tool in the committee's toolbox. But if he's interested in how the teams ought to be ranked and seeded, not how they will be, Sagarin's are precision tools. The RPI is a blunt instrument.

In fact, I came away from this story persuaded that fans do indeed put more stock in the RPI than the committee does, just as the NCAA has insisted for years. Further, I now believe that plenty of thoughtful committee members past and present fear that they've created a monster, and many would love to swap the RPI for something better. Recent members from big and not-so-big conferences alike, men like Big East commissioner Mike Tranghese and Missouri Valley commissioner Doug Elgin, have gone public about their RPI misgivings.

Meanwhile, I'm taking the pledge: I will no longer invoke the RPI in making a case for or against a team. I will respectfully refuse to engage in debate with those who do, and will suggest any of Sagarin's three ratings as an alternative. And I will urge Dr. Joe, as he tabletops his brackets, to do the same.

Because if there were a rating system for rating systems, the RPI would be somewhere between New Hampshire and Savannah State.

SI senior writer Alexander Wolff is the author of Big Game, Small World: A Basketball Adventure. You can contact him at http://www.biggamesmallworld.com.

 

Related information

Stories

Alexander Wolff's Insider Archive

Multimedia

Visit Video Plus for the latest audio and video