The committee that will select and seed the field of 68 for the NCAA Division I men's basketball championship in the coming weeks has a daunting task. Determining who should go to the tournament and where they should be seeded is not simple.
Because the process is so complex, the committee recently began a yearlong study to improve the information at its disposal in the selection process, including an effort to utilize better analytics. I was invited to Indianapolis by the NCAA staff to participate in the initial meeting and discuss which metric or sets of metrics the committee should use. A central question raised was about what the committee values most: Does it care about the best team, or the team that has accomplished the most during the season?
To illustrate the difference between these concepts, fellow attendee Jeff Sagarin presented an extreme situation: Imagine a team that plays only the top 15 teams in the country, and loses each game by one point. In this example, the team has accomplished nothing (it has zero wins, after all), but since its losses have been extremely close and against the best competition, this is probably a very good team.
Indiana follows this pattern to a degree. The Hoosiers are 15-12, but they have lost by five points or fewer to a series of very good teams, including Purdue, Wisconsin and Butler, suggesting that while the Hoosiers have not accomplished a lot this season, they are probably better than their record suggests.
A less extreme and more real example, one that falls on the other side of this season's coin, would be the Maryland Terrapins. Evaluating Maryland via ESPN's separate measures of team strength (Basketball Power Index) and team accomplishment (strength of record) shows a team worthy of considerable debate.
Maryland ranks an impressive 17th in strength of record (SOR), which values the difficulty of every game based on opponent, the location of the game and how rested each team is, among other variables. Even after losing three of five, Maryland sits at 22-5, including a quality road win at Michigan (24th in BPI) and a quality home win versus Indiana (29th in BPI), among others.
Despite those successes, the Terrapins rank 40th in BPI, a predictive measure of how good a team is and how well it can be expected to play going forward based on season-long data. Maryland has risen in the BPI rankings (it was 64th at one point early in the season) but is not a top-rated team by this metric because its wins and losses -- including the margin of those wins and losses -- do not reflect a top-20 team at this stage.
BPI is built to predict future wins and losses, and one of the best predictors of future team strength is scoring margin (or more specifically, a team's adjusted net-per-possession scoring margin). Maryland defeated Michigan by seven points, exceeding expectations and moving up four spots in the BPI rankings in the process. But the Terps also lost to Pitt (58th in BPI) by 14 when they were expected to win by three, and lost to Nebraska (97th in BPI) by two when they were expected to win by 11.
So one of the committee's challenging decisions will concern Maryland. If the committee looks at what Mark Turgeon's team has accomplished this season, Maryland is probably a 4- or 5-seed. But based on how well the Terps can be expected to perform in the tournament according to the more predictive BPI, they should be viewed as a bubble team that should be no better than a 9- or 10-seed, behind teams such as VCU (35th in BPI), Xavier (30th in BPI) and Oklahoma State (28th in BPI). The Terps have accomplished more than all three of those teams and have a stronger strength of record, yet predictive measures of team quality suggest they are less likely to have success in the tournament than that trio.
So what is the committee to do in a case like this?
The first step is to recognize that there are two distinct decisions to be made. Should a team get into the tournament? And if the team is in, where should it be seeded?
As these are different questions, we probably need different metrics to answer them. Very few people would logically argue that Jeff Sagarin's extreme team with no wins should be included in the tournament -- it really did not earn it. When it comes to selecting the field, we should measure what a team has accomplished and use that as a guide for who gets admitted. Since that's what the strength of record metric does, SOR or a similar metric that measures accomplishment on a game-by-game basis should be the metric of choice. That evaluation lands Maryland in the tournament, and the hypothetical winless team is out.
Seeding is a different decision, however. Once the field of 68 has been determined, the seeds should reflect the quality of the teams. We want the best teams to have the top seeds, which requires a predictive measure of team quality such as BPI (Sagarin, Ken Pomeroy and others have their own predictive rankings as well). Using BPI to seed the tournament right now gives Maryland a 9- or 10-seed.
This two-stage process allows teams to earn their way into the tournament without worrying about whether they win by "enough" to satisfy a metric that utilizes scoring margin, but also does not overseed a team just because a few close games happened to go its way.
The committee will consider this and other options over the next year before determining how best to revise the process, but in the end, in some fashion, the NCAA has to address the fundamental difference between best and most deserving.
For more from ESPN Analytics, visit the ESPN Analytics Index.