Tuesday Stats: The Siren Song of the Six Seed

Sturgeon’s law states that 90% of everything is crap. When it comes to basketball message boards, it’s probably closer to 99%. Still, I read more Big Ten basketball boards than I care to admit. Why? Because I’m a basketball junkie.

But I came across something this week that just might fall into that elusive 1%. It was an argument that if your goal is to reach the Elite Eight, you should be happier with a six seed in the NCAA Tournament than with a four or a five, because you get to avoid the one seed for longer. Hm.

My intuition was that this was wrong, because a six seed is going to get a more difficult first and second round opponent, making it less likely that they will even make the Sweet Sixteen, and you have to make the Sweet Sixteen before you make the Elite Eight.

But that intuition mainly comes from an inherent conservatism. If it’s true that you’d sometimes rather be a six than a four, then something with the NCAA Tournament is broken, and whenever something is broken, you get fans and media members who start clamoring for changes to a system that, let’s face it, is pretty freaking awesome as it is. No, damn it, we don’t need to expand to 128 teams, we don’t need to allow six fouls, and we don’t need any more rules that will cause the refs to have to go to the goddamned monitors anymore than they already do.

Fortunately, in this case we don’t have to rely on my intuition and my crusty get-of-my-lawn attitude. Questions like whether a six seed is more likely to make the Elite Eight than a four seed can be modeled probabilistically, which is kind of my job as the BTP stats guy. So that’s exactly what I did.

What I did NOT do was look at historical performance by seed. There are several reasons for that:

• There haven’t been thousands and thousands of NCAA Tournaments, so the small sample size means that it just takes a couple of broken brackets to completely skew the results. You can never count on a broken bracket.
• The average strength of a 4-5-6 seed relative to a 1-2 seed varies by year. I want to look at 2018-19 specifically.
• Any idiot can go on Wikipedia and count up Elite Eight appearances by seed, and I don’t want to be just any idiot.
• Most importantly, the Selection Committee has at times (OK, like almost every year) made big mistakes in terms of over- and under-seeding certain teams. If you can somehow land in a bracket where the eleven seed really should have been a thirteen, or where the three really should have been a five, then hell yes you want to be a six seed in that bracket. That kind of stuff is all over past data. But we can and should eliminate it from our model, because there’s no way to systematically predict how the Selection Committee is going to screw up this year.

For purposes of our model, we are going to assume that the four best teams in the country are one seeds, the next four best are two seeds, etc. That gets rid of Selection Committee bias/mistakes and lets us just focus on the underlying math. I included auto-bids, assuming the best team in each conference won its conference tourney.

The Analysis

Using the current KenPom efficiency margins, I got the following results:

Probability of making the Round of 32:

• Four seed: 82%
• Five seed: 71%
• Six seed: 62%

Probability of making the Sweet Sixteen:

• Four seed: 49%
• Five seed: 37%
• Six seed: 28%

Probability of making the Elite Eight:

• Four seed: 15%
• Five seed: 10%
• Six seed: 11%

Holy cow, the crazy message board people were right! Well, not about the four seed. But the five, yes. By a percent.

The data bears out that if your goal is to make the Elite Eight, you would rather be a six than a five. And the model is a little bit biased because the sixes are all worse than the fives, whereas in reality a team’s fans aren’t hoping that they will actually be worse, but just be under-seeded.

We can model that, too. Coincidentally, one of the five seeds in my model is a Big Ten school, Maryland. As a five seed in their current region, they have an 8.31% chance of reaching the Elite Eight. The six seed in that region is Oklahoma. If I flip Maryland and Oklahoma and rerun the model, Maryland’s chances of making the Elite Eight jump up to 15.38%. Wow!

That’s partly because Maryland was assumed to be in Virginia’s region, and avoiding Virginia is a big deal, because Virginia’s efficiency margin is through the roof. If I move them to Gonzaga’s region instead, their odds as a five seed are 12.94% and their odds as a six seed are 13.26%. That’s a much smaller difference, but it’s still an increase.

Conclusion

It turns out avoiding good teams is a really big deal. I ran a similar analysis for reaching the Sweet Sixteen, and sure enough being a ten seed is better than being an eight or a nine.

That said, a team’s ultimate goal shouldn’t be to make the Sweet Sixteen or the Elite Eight. It should be to win it all, or at least reach the Final Four. To do that, assuming no upsets, there’s no avoiding that one seed. You’ll be happy to note that my model shows that a four seed is more likely to make the Final Four than a five, a five is more likely than a six, a six more likely than a seven, all the way down the line. The system isn’t totally broken.

Don’t hope for a six seed instead of a five seed. That’s dumb. But if you do get the six, by all means log on to your message board of choice and tell everyone that, hey, at least you have a clearer path to the Elite Eight.