Last March Madness, I achieved international fame (not really) for unequivocally predicting that 14-seeded Georgia State would upset mighty 3-seed Baylor in the first round of the NCAA Tournament. And when R.J. Hunter nailed a three with 2.6 seconds left on the clock to cap off a remarkable comeback and down the Bears, I was pretty happy.
But I could have been happier. I had only given readers one game to pick, when there are 67 games that take place in the tournament. So this year I’ve decided to take that prediction schtick to the next level: I’m going to help you pick your entire bracket. Of course, you could go anywhere else to find out how to fill it out. But instead, I’ll be telling you how not to fill it out, because there are plenty of mistakes you can make.
For starters, please, please don’t fill out your bracket by picking the higher-seeded team. While this might seem like a wise statistical choice, the higher seed isn’t always the team with the higher likelihood of winning. For example, 7-seed Oregon State will play 10-seed VCU in the first round. But don’t be confused by the seeding — according to our projections, Oregon State has just a 36.16 percent chance of beating the Rams.
Additionally, even if the chances of a lower seed advancing are relatively low, the chances of all lower seeds losing is much lower. Take the 13-4 matchup, where upsets generally have a pretty low chance of occurring. None of the 13 seeds has better than a 29 percent chance of advancing. But the probability of at least one of those 13 seeds advancing is more than double that, at 58.5 percent, meaning you would expect at least one 13 seed to advance about three out of every five times.
Conversely, don’t pick an 12-seed to upset a 5-seed just because the 12-5 matchup is a so-called “hotbed” for upsets. You hear it every year: “a 12-seed has beaten a 5-seed in almost every year!” But that number doesn’t mean much. The fact that a team is a 12-seed or a 5-seed has almost no bearing on how good the team actually is. Not all 5-seeds are created equal.
Take Baylor, one of the four 5-seeds in this year’s tournament. Although all of the 5-seeds are relatively strong this year, the Bears are the worst of the bunch.
But you wouldn’t see that just by looking at the seeding, and if you arbitrarily picked a 5-seed to lose in the first round, you would most likely choose one of the other three 5-seeds, all of which have much lower chances of losing in the first round.
The same concept applies for dropping very high seeds from your bracket in the Round of 32, which is a common practice among bracketeers. “The numbers speak for themselves,” you might hear. “Only once in the last 17 years have all of the 2-seeds advanced to the Sweet 16!”
But again, resist the urge to pick a team to lose just because of the number beside its name. Consider Michigan State, the No. 2 seed in the Midwest. You might arbitrarily pick it to lose in the Round of 32 because you want to have a 2-seed lose. But that would be a terrible decision, as the Spartans have the second-highest rating of any team in the entire field, by our numbers.
One last point: don’t pick an upset based on the name value of the school. Everyone remembers Florida Gulf Coast’s two miraculous wins in the Tournament three years ago or Stephen F. Austin’s improbable overtime victory over VCU the following year. But neither of those teams looks particularly impressive this year. Keep in mind there’s a lot of player turnover in college basketball, and a team from three years ago is likely entirely different now. This holds true for FGCU and S.F. Austin, which both look relatively underwhelming this year. Find the next big upset school, and don’t get caught up in past performance. It isn’t relevant.
So here’s my suggestion, and how I’ll be filling out my bracket this year: We’ve built a calculator that will tell you the probability of any team in the field beating any other team in the field. Take a look. Plug in two teams that will play each other and use that percentage to determine who you think might win.
If you want to have an extra bit of fun (and be far less biased and more empirical), take that percentage over to a random number generator. Have it generate a random number between one and 100. If the number generated is less than the probability you found in our calculator, pick the favorite to advance. If it’s higher, pick the underdog. Always stay impartial, but most importantly, always pick Wisconsin.
Update: March 16: A typo that could have influenced the meaning of a sentence was updated.
