March Madness Seeding: Problems, Solutions, & Predictions

At the start of the month, I outlined my issues with the NCAA Tournament selection committee’s current process for allocating at-large bids and suggested how it can be amended. In Part 2, I’m looking at how the committee should seed the field once the 68 teams have been set.

The NCAA Basketball Tournament has 36 at-large bids available for teams that don’t win their conferences. The following shouldn’t be a controversial statement: those bids should go to the 36 teams with the best overall resumes from the current season. No team should be able to erase a bad start with a late-season winning streak. Likewise, stumbling down the stretch shouldn’t nullify a great first month or two. Otherwise, what’s the point of early-season games? All regular season games must have some meaning and impact, and a good win or bad loss early should affect whether you get the chance to vie for a national title.

That said, once the 68-team field has been set (36 at-large bids + 32 auto-bids for conference champions), the process for seeding those 68 teams should look at different factors than the selection process. The selection committee should pit the strongest teams against the weakest teams in hopes of saving the best possible matchups for the championship. To accomplish that, teams should be seeded based on their relative strength at the time the tournament starts (which I’ll dub “current strength”).

As a whole, we’ve decided that 68 teams deserve a chance to play for the championship in a single-elimination tournament. Is that the best system for finding a national champ? Debatable. But it’s the (thoroughly entertaining) system we have. Given that paradigm, our next goal should be to create the best possible final/Final Four and, I’ll say again, that means seeding teams according to their current strength.

No one would be happy if the four strongest teams entering the tournament were lumped into the same region. That would ensure only one of those teams reached the Final Four. Likewise, we shouldn’t be content to see five of the top-ten seeds in a single region, while another houses only one. The nature of single-elimination tournaments is that the stronger team won’t always advance; every team has off days. But there’s no reason to stack the deck against the strong teams by loading up certain regions.

In my earlier article, I covered how some statistical metrics – like KenPom and Sagarin – are better indicators of a team’s true strength (and future success) than others – like RPI or straight-up win-loss records. KenPom is much more accurate at predicting outcomes than RPI, and its forecasts end up very close to Las Vegas betting lines. (In case you weren’t aware, Vegas lines are awfully accurate; the entire existence of the sports-betting industry depends on it.)

The NCAA selection committee already recognizes that current strength should play some role in seeding. When there’s a major injury to a key player right before the tournament, a team will often be dropped down a seed-line. That’s why no. 1-ranked Cincinnati, which lost Player of the Year Kenyon Martin, was a two-seed in the 2000 tournament. It’s probably why Oregon, which just lost Chris Boucher in the Pac-12 tournament, is a no. 3 seed instead of a no. 2 this year.

But, looking at how many teams are drastically under or over-seeded every year, including 2017, the selection committee clearly needs to do more. It should be looking at factors like: win streaks entering the tournament, margin of victory, and advanced metrics like KenPom.

The Current System

For the most part, the committee seems to use the exact same factors for allocating bids as it does for seeding. For example, the last four at-large teams to make the field are sent to the “First Four” play-in games as no. 11 seeds. Because they were the last four teams to qualify for the field, they get the four worst seeds for at-large teams. It doesn’t matter if they, say, went 19-3 after a tough 6-5 start to the season and finished top-15 in KenPom, like a certain 2016 Wichita State team.

Looking at their resume as a whole, the 2016 Shockers deserved to be one of the last teams in the field. I repeat: their early losses have to matter for something. But once you decide they deserve a spot, you shouldn’t pretend that they aren’t way better than the other teams that barely made it in.

The committee did just that, however, and sent Wichita State to meet Vanderbilt in the First Four.

Surprise! The Shockers won by 20.

Then they “upset” no. 6 Arizona in the Round of 64 by double-digits.

That setup wasn’t fair to Wichita State. They shouldn’t have had to play an extra game. It took Gregg Marshall, Fred Van Vleet, and Ron Baker some time to put all the pieces together last year. But once they got it figured out, they were downright dominant.

The setup was even less fair to Vanderbilt and Arizona. Vandy should have been matched up with a team that was on its level in the First Four. Zona should have been matched up with a team like Vandy in the next round. Both of those teams earned that much (assuming they were, themselves, properly seeded).

The Problems with This Year’s Bracket

The 2017 bracket was 2016 all over again. Wichita State, which came into the tournament ranked eighth in KenPom and riding 15-game win streak, was given a no. 10 seed. That means the committee thought at least 36 teams were better than Wichita State. Sure, the Missouri Valley Conference didn’t offer stiff competition this year. But the Shockers only played one single-digit contest over that 15-game win streak. They were demolishing teams, including Illinois State, which was one of the first teams left out of the bracket. Just because the MVC was ultra-fine silt, it doesn’t mean the Shockers were any less of a granite-crushing pile driver. But the committee members don’t consider margin of victory when seeding, so reality was lost on them.

No. 7 Dayton suffered for the committee’s mistake. The Flyers received a no. 7 seed after winning the A10 and got “rewarded” with a first-round matchup with Wichita State. Dayton went in as six-point ‘dogs.

Surprise! The Shockers won (64-58; I told you Vegas was good).

Even Gregg Marshall felt bad about it.

The biggest problem with the selection and seeding process is that it’s reliant on top-50 wins. Mid-majors like Wichita State and, to some extent Dayton, don’t have nearly as many opportunities for top-50 wins as mediocre teams in power conferences, especially in the latter part of the season. If they don’t schedule tough out-of-conference games — and also win those games — they can’t make up for it in conference play. And that’s a huge problem because those non-conference games come early in the season when teams are still gelling.

The 2017 Wichita State team did the exact same thing the 2016 team did: stumble early and then piece things together in the second half. The early struggles were expected since the 2017 team lost four-year starters Fred Van Vleet and Ron Baker to the NBA and started a freshman at point guard.

But based on the current seeding system, there was literally nothing Wichita State could have done to improve its standing to make up for the early losses. Their MVC slate didn’t feature top-50 teams; they had no way to prove to the committee that they were more than a 10-seed. Marshall and company shouted as much — at the tops of their lungs — with routine 20 and 30-point wins. Everyone in the country heard … except the committee.

This isn’t a problem that’s confined to Wichita State and the MVC. It just so happens that they’ve been the biggest victim over the past two years. In reality, the top teams in other mid-major conferences suffer the same predicament (see Dayton and VCU in the A10; SMU and Cincinnati in the AAC; Nevada and Colorado State in the Mountain West).

Prediction: the selection committee revises its approach to seeding in the next two years

The selection committee is not filled with stupid people. But it is filled with bureaucrats, of a sort, who feel their hands are tied by the existing rules for selection and seeding.

The general consensus was that the selection committee picked the right at-large teams this year and then dropped the ball on seeding. It’s not the first time the committee has been criticised, and it’s responded to criticism proactively in the past. This year, it brought in Ken Pomeroy, Jeff Sagarin, and their algorithmic ilk in order to fine tune the selection process. That’s going to lead to change in the coming years.

When you couple that apparent willingness to change with this year’s seeding backlash (from big-name coaches like Marshall, Calipari, and Archie Miller) we should be heading for more fine-tuning: eventually, the committee is going to recognize that picking the field and seeding the field require two different approaches.

The former should rely more heavily on overall resume. The latter should factor in current strength, and that means looking at win streaks entering the tournament along with average margin of victory and advanced metrics.

It might not happen next year, but by the 2019 tournament, I expect factors like your record in the last ten to 15 games, margin of victory, and KenPom ranking to play a role in where teams get seeded.

I firmly believe the committee wants to do right by all these teams. It’s just confined by its stated process. The non-stupids on the committee will eventually figure out that they have the power to change that process, and change it they will.