Selection Sunday sparks a raging debate among college basketball fans: Which four teams deserve a 1 seed from the NCAA tournament selection committee?
This question, however, begs another one: How much does getting a 1-seed actually matter?
Two weeks ago, we took at the eight team race for a 1-seed this year. 1-seeds historically perform very well in the NCAA tournament, so there is understandably a lot of media and fan focus on — or rather, obsession with — that seed line. After all, 1-seeds are (in theory) the four best teams in the country and have (in theory) the four easiest roads to the Final Four.
Still, NCAA tournament 2-seeds share a fairly similar story. Is all this obsession over 1-seeds justifiable based on past tournament history? And more generally, just how important is seeding overall?
Is seeding or team strength a better predictor of NCAA tournament success?
Before I started working with TeamRankings, I compiled data about every NCAA tournament team for the last 10 years using kenpom.com. One of the many statistics I collected was a pre-tournament adjusted efficiency rating for every team. The rating is an effective way to measure team strength heading into the NCAA tournament, so I’ll be using both adjusted efficiency and TR’s stats for this post.
As a basic first step, I looked at the correlation between seeds and NCAA tournament wins, and between pre-tournament adjusted efficiency and tournament wins. Since the 2004 NCAA tournament, the r-squared value between NCAA tournament seeding and NCAA tournament wins (excluding play-in games) is .36. That means that if all you knew were the teams’ seeds, you would be able to retroactively explain roughly 36% of the NCAA tournament wins in the 10-year sample.
On the other hand, pre-tournament adjusted efficiency and NCAA tournament wins have an r-squared value of .32, which implies that seeding could actually play a larger role than team strength in explaining tournament success.
Does the NCAA selection committee really do better at projecting a team’s tournament success than advanced stats do? Stat geek biases aside, that conclusion seems hard to believe. Instead, this is probably some preliminary evidence supporting the importance of seeding.
Do the best teams get the best NCAA tournament seeds?
The NCAA tournament selection committee looks at descriptive information (e.g. team “resumes” largely focused on wins and losses) as opposed to predictive information (e.g. efficiency differential stats). As a result, cases can arise like the Florida Gators in 2013. The Gators received a 3-seed despite leading the entire nation in pre-tournament efficiency.
However, 2013 Florida is actually an exception relative to the last 10 years. Take a look at the graph below:
2012 Ohio State and 2013 Florida are the only two teams with a pre-tournament efficiency differential of over 30 to not get a 1-seed in the NCAA bracket. In general, the committee’s top four teams have been in line with predictive metrics.
Another thing to note is that the gap in efficiency between 3-seeds and 4-seeds has not been nearly as big as the gap between 1-seeds and 2-seeds. This shouldn’t come as a huge surprise, since most years there are a couple dominant teams that separate themselves from the rest of the country. It’s much easier to distinguish between those dominant teams and a 2-seed than it is to differentiate a good 3-seed and a (usually not that much worse) 4-seed.
How are NCAA tournament teams of similar strength affected by their seed number?
To draw some better conclusions about seeding, we have to account for team strength. I grouped every 1-seed through 4-seed in the last 10 years into five categories based on efficiency differential: elite, great, very good, good, and above average. This allows us to look at how, for example, “great” teams have performed depending on seeding.
Sample size is still a significant issue even with 10 years worth of data. The NCAA tournament is famous for its variability and we are only working with 40 teams at each seed number. Still, we can look for any striking differences, and for patterns that are consistent across bins:
Notice how “great” and “very good” are the only two classifications with at least five teams at each seed line. In both cases, with the exception of the very small sample of only five “very good” one seeds, there is a consistently downward sloping trend of NCAA tournament wins as seed number gets worse. The 20 “great” 1-seeds have won an average of three NCAA tournament games, but the 10 “great” 4-seeds have won only 2.3 games on average.
Compare this to the trends going down each column. “Elite” 1-seeds have won 4.2 games per tournament while “very good” 1-seeds have won only 2 games per tourney. Sample sizes in some bins are small, but in every case where there are more than a few teams, the expected pattern is present — worse teams win fewer games in the tournament.
It seems clear that both seeding and team strength matter. Still, you can’t look at this data and conclude that a seed difference of one or two is hugely meaningful for a specific team.
It’s important to remember that the dynamics of every NCAA tournament are not the same. The bracket is designed to give the 1-seeds the easiest roads, but that might not necessarily be the case today. It benefits a team (on average) to get as good of a seed as possible, but seeding alone shouldn’t be used to justify your bracket pool picks this year. Even if a team takes care of business in their conference tourney and jumps up a seed line, a bad draw is a bad draw.
Thanks for reading! If you liked this post, we’d appreciate a tweet: Tweet this post