The committee’s mandate, per N.C.A.A. language, is to choose the “best” teams. But how is best defined? Is it the teams with the best résumés from the season, based on the traditional factors of win-loss records against quality opponents? Or is it the teams that are likely to perform best in the tournament, even if they do not have sparkling winning percentages?
In summary, it is a question of results versus prediction. Should the season’s results carry the day? Or should a broader sense of which teams are more likely to advance be the decisive factor?
“That’s the debate that happens every meeting with new people,” said Doug Fullerton, the former Big Sky commissioner, who served on the selection committee from 2010 to 2014. “Some people will raise their hand and say, ‘Are we picking the teams that deserve to be there — did what they were supposed to do — or the teams we think can win?’
“Someone will say, ‘We’re not trying to pick who wins.’ And I’ll say, ‘Well, we obviously are.’ If we weren’t trying to predict who wins, we wouldn’t be seeding.”
This year’s committee chairman, Mark Hollis, Michigan State’s athletic director, tried to explain his mind-set to reporters on Wednesday. “It’s not necessarily the most deserving teams,” he said. “It’s the best teams based upon results.”
Several close observers acknowledged the debate is somewhat one of semantics. But not completely. Imagine, for instance, two teams that have identical records except for five games. One team won those five games by a point each; the other team lost those five games by a point each. The results say the teams are vastly different; a reading of the fine print suggests they are roughly the same.
“Especially when the results are very narrow wins at the end of a game, that tends to be confounded a lot with luck,” said Joel Sokol, a Georgia Tech engineering professor who devised a predictive metric called L.R.M.C. (It stands for Logistic Regression/Markov Chain, though you probably already knew that.)
Sokol added, “The predictive methods — I don’t want to say they ignore who won or lost — but they don’t treat a 1-point last-second win and a 1-point last-second loss very differently.”
Wichita State is an example of a team that was better than its record last season. It started slowly and finished 24-8 against a comparatively weak schedule — nominally the résumé of an N.I.T. squad. But the Shockers mostly blew out their Missouri Valley Conference competition and played close games against the N.C.A.A. tournament teams Iowa, Northern Iowa and Seton Hall. Predictive metrics smiled upon them. The committee decided to unhook the velvet rope for the Shockers, who proceeded to “upset” sixth-seeded Arizona in the first round.
St. Mary’s is a good team to watch as a bellwether for the committee’s thinking this year. With a 28-4 record, the Gaels are hardly at risk of being snubbed outright. But their West Coast Conference competition and their dearth of wins against highly regarded opponents mean a conventional, results-oriented analysis would probably leave them with a disadvantageous, double-digit seeding.
But an estimation that reflected how they played their games — almost always winning by double digits, and with three of their four losses coming against Gonzaga, a likely No. 1 seed — would seem to suggest they merit better placement. L.R.M.C. calls them the 18th-best team in the country and ratings at KenPom.com, an advanced analytics site, have them at 13th. One could argue that anything lower than a No. 5 seed, then, would be a serious slight.
The introduction of predictive metrics, which have been embraced by the N.C.A.A., has only complicated this fundamentally philosophical debate. Because while the N.C.A.A.’s traditional rating percentage index focuses on win-loss records, the new metrics tend to account for margins of victory or defeat, which frequently convey more information about a team.
However new some of the cutting-edge statistics are, the debate itself is an old one, according to Bill Hancock, who was director of the men’s basketball championship for more than a decade.
“We were having the same debate 20 years ago,” he said, adding: “What that came down to would be to ask the former coaches on the committee, ‘Which one do you not want to play?’ That’s another way of saying, ‘Which is the better team?’”
Now, as executive director of the College Football Playoff, Hancock works with a committee that tackles a similarly heated and high-stakes debate — how to pick the “best” four teams based on a season’s worth of games — and seems to have taken a pretty clear stand on prediction over results. In 2014, the football playoff committee slotted two one-loss teams over undefeated Florida State. Last season, it slotted Ohio State (11-1) above Washington (12-1), a conference champion, and Penn State (11-2), a conference champion that had beaten Ohio State that season.
“So many games are decided by momentum, emotion, injuries, the way the ball bounces,” Hancock said, adding, “The debate was over early for us: the best four teams.”
Of course, there is a crucial difference between the two college championships. The football playoff restricts itself to four teams playing two rounds. The fifth-best team is barred from title contention.
A 68-team tournament with six full rounds is a different story. As Sokol noted, with 68 teams in the tournament (and more important, 36 at-large teams), “it’s a virtual certainty that the best team will be included.”
“Then,” he added, “the onus is on them to win.”
Continue reading the main story