Here's how to pick March Madness upsets, according to the data
Dust off your crystal ball because it's time to get ready to make this year's March Madness selections.
Whatever your sweet spot is for predicting upsets in the NCAA tournament, there is one important question that applies to every fan: How many upsets should I pick?
We examined the last 33 NCAA tournaments since the field expanded to 64 teams in 1985 and crunched the numbers for each round. Now we're here to encourage you to make as many upset picks as possible without going overboard.
|MARCH MADNESS SHOP|
First, let's set some ground rules. We defined an "upset" as when the winning team in an NCAA tournament game was seeded at least two seed lines higher than the losing team. There are too many common matchups where the difference in seeding is only one seed line (e.g. the 8/9 game in the first round, the common 4/5 game in the second round, and a 1/2 matchup in the Elite Eight), so it seems to take away from the spirit of an "upset" to say that a No. 9 seed beating a No. 8 seed is an upset.
Using this definition of an upset, in 26 of the last 33 seasons there have been between 10 and 16 upsets in the NCAA tournament. The annual average is roughly 12.7. There have been as few as four upsets (2007) and as many as 19 (2014) but the sweet spot is obviously somewhere in between.
If you're looking for a floor (i.e. "How many upset picks is too few?") then 10 is a good number to try to hit. There have only been four years in the last 33 in which there were fewer than 10 upsets in a single NCAA tournament and 2007 was the only season in which there was fewer than eight.
Here's the full breakdown.
|Total Upsets||12.7||4 (2007)||19 (2014)|
|First Round||6.1||2 (2007)||10 (2016)|
|Second Round||3.6||0 (3 occasions)||8 (2000)|
|Sweet 16||1.6||0 (5 occasions)||4 (1990)|
|Elite Eight||0.9||0 (10 occasions)||3 (2 occasions)|
|Final Four||0.3||0 (23 occasions)||2 (2014)|
Looking at the averages listed above, your number of upset picks should roughly be half as many as there were in the previous round, given both the number of games per round and the quality of teams that advance each round.
So, to answer the first question posed earlier, 20 upsets is probably too many (and six is probably too few). Now, on to an equally important question: Where should those upset picks be made?
Here's a breakdown for the first two rounds of each upset based on seed lines, the number of times such an upset has occurred since 1985 and the percentage of the time such an upset has happened.
|First Round Upset||Frequency||Percentage|
|No. 10 seed over No. 7 seed||51||38.6%|
|No. 11 seed over No. 6 seed||49||37.1%|
|No. 12 seed over No. 5 seed||47||35.6%|
|No. 13 seed over No. 4 seed||26||19.7%|
|No. 14 seed over No. 3 seed||21||15.9%|
|No. 15 seed over No. 2 seed||8||6.0%|
The individual matchups for upsets in terms of seeding become less common after the first round, based on the results of the first-round games. But if we combine some of the seed lines within individual four-team pods (e.g. a No. 7/No. 10 seed upsetting a No. 2 seed, a No. 6/No. 11 seed upsetting a No. 3 seed, etc.), it'll provide a clearer picture for how many upsets you should pick in your bracket.
Teams seeded No. 7 or No. 10 have upset a No. 2 seed 41 times in the last 33 years (1.24 per year), and No. 6/No. 11 seeds have beaten No. 3 seeds 43 times (1.30 per year), so history tells us you should probably pick at least one No. 2 seed and one No. 3 seed to lose in the second round. Seventeen times a No. 1 seed has lost in the Round of 32, which means that roughly once every other year a No. 8 or No. 9 seed knocks off the top seed in its region.
|Second Round Upset||Frequency|
|No. 6 seed over No. 3 seed||28|
|No. 7 seed over No. 2 seed||23|
|No. 10 seed over No. 2 seed||18|
|No. 11 seed over No. 3 seed||15|
|No. 8 seed over No. 1 seed||12|
|No. 12 seed over No. 4 seed||12|
|No. 9 seed over No. 1 seed||5|
|No. 13 seed over No. 5 seed||3|
|No. 14 seed over No. 6 seed||2|
|No. 15 seed over No. 7 seed||1|