NCAA Tournament: Selection Committee Over-Emphasizes Big Wins in Regular Season, Again Proves Foolhardy in March

I’ve railed against the RPI–the Ratings Percentage Index–for years. I do think that the RPI has many issues. But I’m beginning to think that the biggest problem when it comes to the NCAA Tournament seeding isn’t the RPI, but rather the thought process that leads to rewarding big wins (or rather, punishing the lack of them) in the regular season when a team’s resume is otherwise worthy.

Jason Whitlock had a conspiracy theory when it came to Wichita State being paired up with Kansas yesterday.

Wichita State finished season ranked 14-13 depending on poll. Kansas 10-11. But KU 2 seed WSU 7 seed. Bracket manipulation for attendance.
— Jason Whitlock (@WhitlockJason) March 22, 2015

There’s a much easier explanation, though. The committee flat out gets it wrong when they over-emphasize “big wins” as the outcome-determining factor of seeding. Wichita State isn’t the first case, and won’t be the last unless they change the philosophy.

The RPI, the national rankings, and analytical rankings like Ken Pomeroy all were in basic agreement on Wichita State. They were 14th in the RPI and 14th in Ken Pomeroy entering the tournament as well. Why dropped to a 7-seed? It wasn’t to set up a matchup, as the Committee has done this before. It was because Wichita State only beat Northern Iowa. So, the committee went against basically every other shred of evidence and seeded them as a 7-seed.

In Wichita State’s case, they can only play who is on the schedule. The RPI was in agreement with other measures: Wichita State was a top 15 team. Nope, because “who did they beat?”

But we’ve seen a similar bias be the committee’s biggest bug-a-boo when it comes to seeding errors with big conference schools. Michigan State was within Pomeroy’s top 20 after their run to the Big Ten title loss in overtime to Wisconsin. They moved up to 23rd in the RPI. They looked like a 5-seed, and people reacted accordingly when the brackets were announced, as many saw Michigan State as a very tough draw and potential Elite Eight or Final Four team out of a tough region.

Why were they seeded lower than where both the rankings, and the RPI, and the analytical measures had them? Simple, they went 1-7 against the RPI Top 25, with the best win being Maryland at a time the committee was hunkered down already preparing the bracket. They lost those top games in the regular season, and they got seeded down.

Kentucky last year as an 8-seed was in the same boat. They were 1-5 against top teams in the regular season, and couldn’t beat Florida in three tries. So, a team that was #17 in the RPI, #17 in Ken Pom entering the tourney, and passed every eye test in terms of talent was made an 8 seed.

The Sweet Sixteen members were a combined 49-63 in games against the RPI Top 25. Throw out the three #1 seeds to advance, and the other three were 32-59 against the RPI Top 25 in the regular season.

The RPI may already overvalue the top conferences. The committee then takes that info and over credits those teams even more. The Big 12 and Big East were the two highest-rated conferences. The teams in those two conferences got collectively seeded higher than even their RPI’s by circular reasoning. Teams get credit for “good wins” without getting dinged as much for losses. Every Big 12 team beat each other, giving them all a bunch of “good wins.”

I ran the correlation coefficient between NCAA tourney seed and RPI (assigning a 14 seed to all in the top 64 in RPI that didn’t get an at-large). It was +0.84. Pretty strong. But assign a simple system of +2 for a top 25 RPI win, and +1 for a top 50 win, and add those up and average those “best win” rankings with the RPI, then run the correlation?

+0.94. If you are not familiar with correlation coefficients, that’s as close to matching up as you can get.

Look at win percentage in top games (rather than just wins, which favors the top RPI conferences because of raw opportunity), and the correlation actually goes down to +0.74.

The Committee favors wins, but doesn’t seem to account for losses all that much, nor do they account for a team having less opportunities but a similar win percentage in big games as a team from the #1 conference.

That may not seem important, but this year was no outlier. Over the last eight tournaments, the #1-rated conference in RPI has sent 21 teams as at-larges who were seeded #6 or worse. One- one! — of them reached the Sweet 16, and none the Elite Eight. That’s a lot of misses because of greatness by association. (and that one was Marquette in 2011 as a #11 seed over fellow conference member Syracuse).

On the flip side, while those conferences are being over-seeded by association, the Wichita State’s, Kentucky’s, and Michigan State’s will continue to be under-seeded unless the Committee sees their systematic error. When the Committee ignores basically every other indicator (national rankings, the RPI rankings, and Ken Pomeroy or Sagarin) and seeds a team way lower, well, they usually miss badly.