20.3.17

NONE OF US IS AS DUMB AS CONSENSUS.

Doggone it, I really wanted to lay off of the Wise Experts for a day or two, but this one is too juicy to pass up.

During the college basketball tournament, the business types get their knickers in a twist over how much time is lost at setting up office bracket pools (or perhaps the fillip to morale the office pools provide) and I was guilty of teasing students, this time of year, "You know, if you devoted half the time you put in on your brackets to indifference at the margin conditions ..."

And if you aspire to bigger things, there's the ESPN Tournament Challenge.  As of this morning, out of over eighteen million entries, all of eighteen have the round of sixteen correct.  It is likely that substantially more than eighteen will have the round of four correct, and a non-trivial subset of those will name the team that cuts down the nets.  (But that will be neither last year's winner Villanova nor Duke from the year before that, and Connecticut's men have been cheering their women classmates for some time already.)

But the failure of more contestants to identify the round of sixteen is a signal of a system failure.  The Wise Experts are up front about rigging the competition, something that promoters of wrestling have long been accused of.  As ESPN's Eamonn Brennan has it, "It's about doing what the bracketing principles and procedures are supposed to do, especially with top seeds, teams that have spent months earning their spots. It's about rewarding those who deserve to be rewarded."  That is, the teams the Wise Experts would like to see meeting each other either in the regional finals or in the round of four get to go through the weakest opponents first, while the middling-strength teams get to test their mettle against each other in a game or two before they get to take on the deserving rich.  And sometimes there's a Cinderella story, although the folks who rig the system are getting better at Cinderella-proofing (or is it George Mason-proofing?) their show.  In 2007, for instance, the round of sixteen comprised no worse than a five-seed.

This year, the Wise Experts whose job is to rig the system started releasing their rankings about ten games into conference play, signalling their likely top sixteen teams. We had, for instance, Purdue somewhere in the initial top sixteen, Wisconsin outside that list, and Michigan seems like a dream to me now.  Those are the three Big Ten teams still playing.  Wisconsin entered the tournament as an eight seed.  Former players, who had the current starters developing as reserves, were skeptical of that seeding before a tip-off occurred.  "Minnesota, one of six other Big Ten teams in the tournament, received a No. 5 seed. The Golden Gophers finished one game behind Wisconsin during the regular season, lost both meetings to the Badgers and were eliminated in the quarterfinals of the Big Ten tournament."  That five can often be a kiss-off, as the twelve beating the five is frequently a good bet; then the expected outcome is the four gets through to the round of sixteen.  Minnesota was the first five out.  "Wisconsin plays Virginia Tech in the first round on Thursday. If it wins, it’ll likely face defending national champion and top overall seed Villanova on Saturday."  Villanova performed as expected against Mount Saint Mary's.  Then they had to play Wisconsin.  Oops.
In Buffalo, unlike in Salt Lake, the No. 8 seed in question was not [Northwestern] making its lovable first NCAA tournament foray but a roster whose seniors have been to two Final Fours and three Sweet 16s and played in 15 NCAA tournament games in the past four seasons -- the most tourney-tested group of players in the sport. In Buffalo, unlike in Salt Lake, the No. 1 seed's path to the second weekend went through one of the most underseeded teams in the 2017 NCAA tournament field.

Yes, Wisconsin was underseeded. Throw out the Badgers' past accomplishments (the selection committee certainly does), and there remains no actual basketball explanation for why Greg Gard's team was seeded where it was. The Badgers entered Selection Sunday 25-9 with a 12-6 record in the Big Ten -- same as Maryland, a No. 6 seed, and one win better than Minnesota, a No. 5 seed that Wisconsin beat twice.
That's Mr Brennan again, and it's the data-driven obsession of sports pundits that deprived East Coasters of the expected Villanova - Duke showdown in the attic of Penn Station.
Why did this happen? Because even as every coaching staff tracks its per-possession performance and Las Vegas builds books based on advanced analytic projections, the people responsible for deciding how the sport's most important competition is structured can't be bothered with all that much more than the RPI [a performance index -- ed].

Wisconsin's RPI was 36. Its nonconference strength of schedule -- which is based on RPI -- ranked in the low 300s. Its "best" wins -- which is to say "best" according to the RPI -- included only two against the top 25.

If you live in the selection committee's world, it isn't hard to understand how a team with Wisconsin's résumé could end up playing the top overall seed on the first weekend of the tournament. If you live in the real world, it's impossible to fathom.
On the other hand, if the Wise Experts got it right, you might give yourself a good chance of winning the Tournament Challenge simply by picking the picks of the Wise Experts and their algorithms.  And the winning team would pay Hillary Clinton a visit in the White House, but I digress.

But the Wizards of Smart are going to double down on Applying Expertise, perhaps with Better Index Numbers.
That's why the National Association of Basketball Coaches asked the committee to join the rest of the sport in the glories of modernity and why the genuinely smart, often forward-thinking folks at the NCAA responded by summiting with some of college basketball's best statistically inclined minds. It's why a new metric might soon replace the RPI -- maybe as early as next March.

Because the bracket could be better. Because it should be. Because days such as Saturday, when the top overall seed faces a team such as Wisconsin, shouldn't happen -- not this early, anyway.
Put another way, a committee with better information ought to have a better chance of rigging the system so as to dispose of the champions of obscure conferences and the upper middle finishers in the power conferences more effectively.  They're unlikely to consider other pairing approaches, such as the Swiss system used in chess tournaments (they could avoid a challenge the tournament director faces by allowing the higher-rated team to wear its home uniforms, there being no first-mover advantage to contemplate) or even partially random approaches.  Perhaps pick the top sixteen teams, then fill in the pairings for the rounds of 64 and 32 by drawing names.  That might be the end of those Tuesday "first four" games, two of which generally involve teams that earn the honor of crash-test dummy, but that wouldn't be all bad, either.

I suspect, though, the Wise Experts will set things up in such a way as to make the seedings better predictors of the outcomes in the early rounds, even if that annoys the people on Tobacco Road.
When the committee released its initial rankings in February, it regarded precisely zero Big Ten teams among the top 16 seeds. Ultimately, none received any seed better than Purdue’s four in the Midwest. Faced with skepticism all season, the Big Ten has mounted a powerful rebuke in the bracket.
The Atlantic Conference fans? In the same position as Lyndon Johnson's "best and brightest," all of Hillary Clinton's position papers, and the Atlanta Falcons. "Duke is a standard bearer and was viewed as a serious title threat. Ditto Louisville. In its worst visions, the ACC could not have imagined a Sweet 16 with both Mike Krzyzewski and Rick Pitino cast as observers." Yes, and the Smart People thought they had a matchup for the ages of Duke at Villanova in New York.

Wisconsin?  Recruit, develop, season the young players, and prepare for the next 40 minutes.  There's beer in the cooler, whatever the outcome.

No comments: