What Happened Dave?!!!




UMBC Upset Confounds All Mathematical Models

Did you see our post about March Madness last week? Author/Engineer Dave Martin used data and PTC Mathcad to show that we can always expect upsets during the first round of the NCAA men’s basketball playoffs.

 If you’re not familiar with March Madness, here’s a 2-minute overview from SB Nation.

In fact, he found that more than 25% of lower-seeded teams beat a higher-seeded team in the first round. So, for example, a #9 team might beat a #8 team. Not an earth-shaking outcome, right?

That said, his data also showed that a #16 team never, ever beat a #1 team in that first round. Not in the entire history of the tournament. Okay, also not all that surprising.

And then the unthinkable happened. #16 UMBC beat #1 Virginia this year.

“Dave Martin has some explaining to do,” said my colleague David Newman in marketing.

Indeed. I found this note from Dave in my inbox on Monday morning:

Cat,

History was made on Friday night. For the first time in the 34 years of the modern 64 team field, a #16 team defeated a #1 seed. Not only that, the University of Maryland, Baltimore County (UMBC) "Retrievers" thoroughly dominated the University of Virginia, who many expected to win the National Championship. As they say, "that's why they play the game." People will be talking about this upset for decades.

There were 9 upsets in total First Round play, slightly higher than the average 8.063 per year. Anecdotally, the number of upsets tends to see-saw each year above and below the average. Last year there were only 6 First Round upsets. I thought there would be more than the average this year, but I never expected a #16 to take a top seed.

Seed breakdown:

  • There were 2 upsets of #4 teams (almost 3, but Gonzaga pulled it out in the last minute). At 50%, that's significantly higher than the historical average of 19.5%. 
  • Two #6's also fell (historical average 35.9%). 
  • One #7 lost to a #10 (historical average 39.1%).
  • A whopping three #9 teams lost to the #8. Normally that happens almost half the time (48.4%), but the seeds are so close to each other, it's not shocking.

The West and Midwest regions only had one upset a piece, but 3 were in the East and 4 were in the South.

It will be really interesting to watch the repercussions this year. A #2 seed loses on average every 4 years, but this UMBC upset is unprecedented and confounds all mathematical models. I'll certainly be tuning into the match-up of two upset teams Sunday night (the Retrievers play #9 Kansas State), because a #16 team reaching the Elite Eight would simply be mind-boggling.

Dave

So I asked Dave, what’s the lesson here? Statistics sometimes lie? That the numbers can deceive us? Here’s his reply:

It proves things like the Gambler's Fallacy, and a little more esoterically, Quantum Mechanics (anything that can happen, will happen).

Most of all, it shows that there are certain things that PTC Mathcad can't take into account - like heart, and those young men had it Friday night.

Dave