A look at picking first round upsets

Upsets. They're a huge part of the NCAA Tournament's popularity. As each person across America fills out their bracket, they pour over each matchup, trying to figure out where each of the upsets will occur. After all, that's where bracket pools are won and lost. Anyone can pick chalk. The winners are the ones who can correctly predict when and where the upsets will occur. If you correctly pick enough underdogs, you'll be the only one in your pool laughing as everyone else curses their busted brackets. Of course, that's easier said than done. They wouldn't be upsets if it was easy to see where they would happen. Still, after the first (or second, whatever) round of the NCAA tournament ends, we know one thing for certain: the landscape will be littered with higher seeded teams, laying on the floor with a look of disbelief, trying to figure out how that undersized team with a mediocre record could possibly have ended their season.

But just how often do these upsets really happen? Are they really so prevalent, or are they so memorable that they're the only games that stick out in our minds, causing us to believe they're happening more often than they really are? In effort to answer this question, and hopefully pick up some helpful tips on filling out a successful bracket, I went through every Round of 64 game from the last ten years to look for trends.

First, you can see that the number of upsets occurring each year stays fairly consistent. Aside from the especially chalky 2007 tournament, there have been between 7 and 10 upsets by seed in each of the last ten tournaments. Over that span, the favorite (team with the better seed) has gone 235-85, giving them a winning percentage of .734. That means in 73% of games, there's no upset. It also means that roughly one out of every four first round games you watch will end up with the underdog on top.

We all know that no 1 seed has ever lost to a 16 seed, and as a result fewer people watch these games. We've all heard about the famous 5-12 upset, and these games tend to be more popular. But in which games are the upsets really happening?

You can see that 1-3 seed games are rarely going to feature an upset. Across the past 10 years, the 1, 2 and 3 seeds are a combined 112-8 in these games. So if you want to pick an upset, avoid these games, because they've happened just 6% of the time. Things get more interesting on the four line where the winning percentage drops to exactly 70%. The odds of an upset here still aren't great from a gambler's perspective, but it's fascinating to note that in a game where the disparity in team quality should be as great as in a matchup of a 4 seed and a 13 seed, that an upset is going to pop up three out of ten times.

Things really get crazy starting at, as you might have guessed, the 5-12 matchups. They aren't quite coin flips, but they aren't too far off. It should be noted that the 5 seeds have only done one game worse than the six seeds across this span, which takes some of the luster off the fabled 5/12 matchups. There's no substantial difference in the rate of upsets across seeds 5 through 8, either. It was interesting to me that the 8 vs 9 games, which should really be a virtual coin flip, still featured the better seed winning 55% of the time. Even down to the last seed, the higher rated team is still winning more than half the time.

So what should your first round picks look like?

Based purely on the data I found over the last ten years, I came up with some recommendations for putting together a bracket that jibes with past tournaments' results

• Don't pick seeds 1-3 to lose in the first round. Unless you're the kind of guy who walks up to a roulette wheel and throws down $100 on 17 because you "have a feeling" about that number. Your odds of finding an upset here are about as good as two spins of roulette
• Don't load up on upsets on any one particular line. No seeding line 1 through 8 has gone 0-4 in the first round (in the last ten years, at least), and no seed 1 through 4 has had a losing record.
  
• There's nothing special about 12 seeds beating 5 seeds. They haven't been any more likely to win than a 9 seed during this span. Seeds 9-12 have been nearly dead even in their likelihood of pulling off an upset, so don't load up on 12s and ignore the 10s and 11s
• Though the higher seeds are going to win the majority of the time, but you can be fairly liberal in picking upsets once you get past the 4 line. Obviously they'll need to be the right upsets if you want to win your pool, but going straight chalk isn't likely to win either. In games featuring a 5, 6, 7 or 8 seed, the better seed wins just 57.5% of the team. So if you go straight chalk on those lines, you aren't doing yourself much more of a favor than you would by flipping a coin. If you feel good about an upset on those lines, go ahead and pick it