There are few things more satisfying in March than watching a huge upset in the first round and being able to brag to your friends that you called it.
But picking upsets carries an inherent risk. They’re upsets for a reason. Get the right ones, and you’ll look prescient. Get them wrong and you can bust your bracket before the first weekend is over.
So, how good are players at picking upsets, and what can we learn from their trends?
We looked at hundreds of millions of picks over the past eight years in our Capital One Bracket Challenge Game to find out.
Here’s how people have picked over that span, compared to the actual results in first-round matchups over the 34 years since the tournament expanded to its modern format in 1985:
| Upset | % picked to win | Actual winning % | Difference | Percent error |
|---|---|---|---|---|
| 9 over 8 | 50.09% | 50.00% | -0.09% | 0.00 |
| 10 over 7 | 44.68% | 38.24% | -6.45% | -0.17 |
| 11 over 6 | 30.88% | 37.50% | 6.62% | 0.18 |
| 12 over 5 | 20.99% | 34.56% | 13.56% | 0.39 |
| 13 over 4 | 11.00% | 20.59% | 9.59% | 0.47 |
| 14 over 3 | 7.71% | 15.44% | 7.74% | 0.50 |
| 15 over 2 | 3.32% | 5.88% | 2.57% | 0.44 |
| 16 over 1 | 2.07% | 0.74% | -1.33% | -1.82 |
The percent error represents how good people are at calling upsets in specific matchups. The closer a percent error is to zero, the more spot-on the picks are. Negative numbers mean the upset is picked too often. Positive numbers show an opportunity where upsets aren’t picked enough.
BRACKETS: Was your 2018 bracket above average?
One thing from the data stands out right away: People pick 16 seeds to win way, way too often. You can expect this problem to get worse this year. After UMBC proved it was possible, more people are going to fall into the biggest trap in bracket picking — eliminating a 1 seed in the first round.
A 1 seed is by far the most likely team to make it to the Final Four. It’s really not even close. Of the 136 teams to make a Final Four in the modern tournament, 56 of them have been 1 seeds. The next highest count comes from the 2 seed, which has 28 appearances. And as a reminder, points awarded for successful picks double every single round. Picking a correct team to win the Elite Eight and make the Final Four is worth 8 points. Picking a correct team to win a first round game is worth 1 point. So even if you are as lucky as one of the 2 percent of brackets that picked UMBC to beat Virginia last year, you’re making a huge mistake by eliminating the team with the highest expected value for the entire tournament in the very first game. If you’re wrong, you eliminate a team that has the best chance of any seed to pick up 15 points in your bracket by making the Final Four just for the opportunity to get 1 point in the first round.
Still, in the eight years we’ve run this bracket game, 16 seeds have been picked to beat 1 seeds in 2.1 percent of their matchups.
MORE TIPS: Here's how to pick March Madness upsets, according to the data
The biggest underestimate here is in the 14-3 games. While the 14 seed has only won 21 games over the 3 seed (15.4 percent), they win twice as often as they are picked to win (7.7 percent). Still, we're not necessarily saying you should pick a 14 seed. In fact, 14 seeds have only won a game in 18 tournaments. But it is a lot more likely than people seem to think.
If we’re talking about actual advice here, we’re not picking 12 seed upsets nearly enough. In the 34 years of the tournament, 12 seeds have won 47 of their 136 first-round games (34.6 percent). That averages out to approximately 1.4 wins per year. In fact, there have only been five years since 1985 without at least one 12 seed winning in the first round (1988, 2000, 2007, 2015, 2018). That means that at least one 12 seed has won a game in 85.3 percent of all tournaments. But 12 seeds are only picked to win 21 percent of their games in BCG brackets. They’re the only seed that wins at least 1/3 of its games that isn’t picked to win even 1/4 of the time.
Naturally, people are pretty spot on with the 8-9 game. That matchup is evenly split at 68 games a side through 136 matchups, and 9 seeds are the ever-so-slight favorites in brackets, picked to win 50.09 percent of the time. As far as seeding goes, that matchup is essentially a toss-up.
If we can leave you with one piece of advice from this data, it's this, which we may or may not have mentioned before: Don't ever, for any reason, pick a 16 seed to win for any reason ever, no matter what, no matter where, or who, or who you are with, or where you are going, or where you've been… ever, for any reason whatsoever.
