Illinois St. left out of the field completely. Monmouth not even considered. Princeton needing an automatic bid. Wichita St. a 10 seed. Thanks to the NCAA Tournament selection committee’s focus on RPI derivatives (not even the metric itself) the accomplishments by mid-majors are consistently undervalued.
Ken Pomeroy laid out a bunch of really good reasons why in his piece on Slate on Monday.
And while RPI broke team selection and seeding, analytics can help fix it.
How should we select teams?
It turns out when you put a committee of people together to make a group decision that they often have different visions of what makes a team the “best” or “most deserving.” It could mean:
- The strongest overall resume
- The most quality wins
- The team that has the best chance of winning the tournament
Right now the committee uses RPI, or some derivative of RPI, to answer all of those questions. But the RPI was developed in the 1980s. Over the last 30 years analytics has dramatically improved, and there are different tools that could help us answer these questions in a much more precise way.
One of the best ways to help mid-majors get a fairer shake in selection and seeding would be to find a metric that puts their entire schedule on the same footing as that of a power conference team. Illinois St. played 12 games against teams that had an RPI of 200-plus this season and unless they lost those games, which the Redbirds did once, the results are basically ignored. Another bubble team, Vanderbilt, played just three of those games. (Though the Commodores also lost one at Missouri.)
Half of those games Illinois St. played against 200-plus RPI teams were on the road, most of them during Missouri Valley Conference play. Would a bubble team normally win all 12? Only 10? There are metrics that attempt to figure this out.
Wins Above Bubble (or WAB) is one such way of empirically measuring a bubble team’s “resume.” Imagine the Platonic ideal of a bubble team. The idea of WAB is figuring out how much better or worse each team would’ve done versus that bubble team. Seth Burn has written extensively about WAB on his site. It turns out a derivative of WAB, Bubble Percentile, is the best way of evaluating how a bubble team would’ve done against a similar schedule. Seth laid out how WAB would’ve seeded the field and Wichita St. would’ve be a 6 seed, Illinois St. would’ve been in as a 9 and Michigan St. would’ve been the last at-large. The first four out would’ve been TCU, Monmouth, Georgia and Vanderbilt.
Maybe you’d like your resume metric from a large, sport-centric behemoth? If that’s the case than ESPN’s Strength of Record is for you. Strength of Record, which is based on BPI, has Illinois St. with the 45th best resume in the countrym in front of bubble teams such as Marquette, Michigan St. and Vanderbilt. (It’s worth noting that Michigan St., which received a 9 seed really didn’t have a great resume.)
It’s possible to judge a schedule holistically. The committee should start doing just that.
The good wins
What if instead you want to know which teams beat the best teams. That is why the committee uses a team’s wins against the Top 25 and Top 50 of RPI as one of its main criteria. But those numbers are so devoid of context as to contain little meaning.
A fair metric would consider a few things:
- The site of the game (home, road, neutral)
- The overall record (and not just wins)
- A better judge of team quality
All of these things could be accounted for using something like a team’s record in Tier A and Tier B games on KenPom. Last January Ken Pomeroy added a marker for these two tiers of games to the team pages on his site. Here’s how he defines both: “A game in Tier A represents a top 50 opponent adjusting for the location of the game, and Tier B is the same concept for a top 100 opponent.”
Illinois St. went 1-3 in games against Tier A and 4-1 against Tier B. That 5-4 record against top 100 teams looks quite different from the 2-4 record that the Redbirds had against RPI top 100 opponents. Why? Because road wins against Loyola (Ill.), Southern Illinois and Missouri St. are now included when evaluating a team. Those are hard games to win!
Major conference teams will still play more of these types of games. Michigan St., for example, went 4-12 in Tier A games and 5-1 in Tier B games, for a combined record of 9-13. Wins at Nebraska and on a neutral court over Penn St. are now included, appropriately, in their overall resume.
The best team
Of course if all you want are the teams with the best chance of winning the NCAA Tournament, that’s easy! Pick your favorite flavor of scoring margin adjusted metric: KenPom, Sagarin, ESPN’s BPI, Massey, T-Rank. Each of those metrics is telling you who they think are the “best” teams in college basketball if they had to go play a game tomorrow.
To me though this feels like a devaluation of the regular season when picking the teams that deserve to be in the bracket. Even the people who created these metrics would probably shy away from using a straight run down the list to select the bracket. But they’re certainly available to consider. And Ben Alamar wrote a good piece on ESPN about why maybe the committee should use these metrics when seeding. (The gist: A mis-seed of Saint Mary’s hurts not only the Gaels but the rest of that bracket.)
No one analytic is the magic bullet and there is always going to be some bit of human decision-making going on in the commitee room. This is about replacing the giant sledgehammer of RPI with the hammer, screwdriver, and wrench of analytics. And there have been discussions about doing exactly that! (Maybe…) The job will get done either way, but the new way is more precise. Hopefully we can all agree that’s a good thing.