I said that I wasn’t going to write another post before Selection Sunday. Then “Around the Horn” had a debate on whether getting a one seed was all that important. And later in the night, some other analyst proclaimed that seeding no longer mattered in the dance because of the parity in college basketball.
Even Ken Pomeroy wrote a piece last year saying that seeding didn’t affect a team’s chances to win the dance. The line of reasoning was that efficiency ratings were what they were and the log5 formula probabilities of winning the tournament would be roughly the same whether a team was on the one, two or three line. Ken rightly pointed out that the most important rounds in determining whether a team would win the tourney were the Final Four and championship game…and that the ultimate champion would have to play great teams no matter what. He also stressed that a team’s chances of winning the dance were affected more by the difficulty of its region, as measured by efficiency statistics.
All of that is well and good if you’re trying to project the ultimate champion. But asking whether seeding matters in predicting the tourney winner is a very different question than asking whether seeding matters at all in tourney outcomes. I’ll agree that a team’s overall efficiency rating is more important than its seed position in determining its chances of cutting down the nets. But I don’t agree that seeding doesn’t matter in earlier round outcomes.
Let’s face it: seeding prescribes a certain path through the bracket. And that path is easier for one seeds than it is for twos, easier for twos than threes, and so on. Not only that, but there’s a significant degradation in the quality of teams seeded lower than 13. That’s because the teams getting automatic bids from small conferences aren’t actually the 52nd through 64th best teams in the tourney. One seeds don’t play the 61st most efficient team in the tournament; they play more like the 195th best team. It’s the difference between playing West Virginia and Albany. On the other hand, five seeds actually do have to play the 45th best team in the country. Maybe that’s why they struggle so much.
But let’s set aside efficiency stats for a moment. Here are the average seed gaps of the games that one through five seeds have to play in the first four rounds of the dance:
We all know the seed differences in round one. But those differences persist throughout the first four rounds. In the second round, one seeds are playing teams 7.5 seed positions away, while two seeds are just 6.2 seed positions from their opponents. Overall in round two, a one-seed difference means about a 1.5-seed gap in the quality of your opponent. And that gap gets bigger in the Sweet 16. One seeds still enjoy a sizable 5.2 seed gap over opponents, while two seeds have tighter games with only 3.3-seed gaps. Three seeds face near toss-up games, while fours and fives are facing higher seeded opponents (usually one seeds—don’t tell me seeding doesn’t matter between a three and four).
I’m right with Ken on the argument that seeding doesn’t matter in determining the ultimate tourney champion. But it does matter in how far some teams progress. Let’s revisit some efficiency numbers. There have been seven teams since 2004 (that’s as far back as I have KenPom stats in my database) who should’ve gotten a one seed based on efficiency numbers—but who wound up getting a three or four seed. If seeding didn’t matter, you’d expect these teams to play like other one seeds, averaging 3.35 wins per dance. In fact, they averaged only 1.29 wins per tourney, more than two wins worse. Not only that, but their average PASE value—even for their actual seeds—was an underacheiving -.336.
The Verdict: Maybe a small difference in a team’s seeding—dropping, say, from a one to a two—doesn’t have a significant impact on its overall chances of winning the tourney. But seeding does matter to a team’s tourney fate, particularly in the early rounds, where the bracket prescribes markedly different paths to the Final Four.
There. I’m glad I got that off my chest. Back to the stats sheet.