Matt’s Musings | NCAA Tournament Data Study, March 16


If you’ve been reading this article series since it’s come out, you’re probably expecting a bit of NFL or PGA data to comb through. Given that March Madness is set to tip off shortly, though, I thought I’d look through some historical tournament data and point out some things that could be of use while filling out the first round of your bracket. Full disclosure: I’m not an expert in college basketball, but I do like looking at unfamiliar data and seeing if I can pull out any base rates that could prove valuable.


Historical Results

My starting point when trying to decide what I should do for a bet is to look at base rates. This is especially true in a sport like college basketball where I don’t have true domain knowledge or expertise. My jumping-off point is always base rates. If I’m not confident in my ability to model something, I look at what has historically happened to make sure I’m making informed decisions. 


Description automatically generated

Strategy for the 1-4 Seeds

It’s awesome to be able to say “I picked that ‘2 over 15’ upset,” but you’re likely being too cute by making that decision in your bracket. If you think you can reliably pick a 6% outcome, more power to you. Given that we have four quadrants of the bracket, I wouldn’t pick an upset of the 1-4 seeds under any circumstance. If that’s the reason I don’t cash in my bracket, so be it. If you’re picking multiple upsets in this range, I think you’re likely punting off money in your contest. Yes, it’s awesome to tell your friends about that random upset you picked. But, I think it’s more awesome to actually … you know, win money at the end. 


Strategy for the 5-8 Seeds

This collection of data goes back through 1985, and there’s a bit of a jump in upset rate here. The 5-7 seeds are all upset between 36-39% of the time with an average score differential of 2-4 points. Those are closer games than I’d anticipate. There’s a lot of focus on the 5 vs. 12 upsets, but since 2015 that’s actually the worst upset to pick out of this group.


Description automatically generated

If just basing it off historical base rates, you’re much better off picking one or two upsets of the 6-8 seeds. There’s also some game theory that comes into play there, as well, since there is so much publicity for the 5 vs. 12 narrative. Again, I’m just a casual observer of college basketball and bracketology, but even I know how much that’s talked about. 


I’m using a few hard and fast rules for filling out my bracket this year. 

  • No upsets of the 1-4 seeds in the first round
  • No 5 vs. 12 upsets at all.
  • Picking one upset each of a 6th, 7th and 8th seed 

It’s not flashy, and you likely won’t have anything super contrarian to brag about. But if you can