When Will a 1 Seed Finally Lose In the First Round?

And Which School Will Be the One to Blow It?

14 seed Mercer upset Duke in 2014

This weekend four teams will move one step closer to cutting down the nets and moving on to this year’s Final Four in Phoenix. Because there have been few big upsets so far in the tournament, save Wisconsin’s upset of 1 seed Villanova and South Carolina’s win over 2 seed Duke, fans should get some quality games that will answer some of our questions going into the tournament. Can Xavier be the just the third 11 seed to advance to the Final Four (George Mason, VCU)? Is this the year stronger-than-ever Gonzaga, appearing in their 19th straight tournament, finally wins a national championship? One tournament question that’s always on the minds of fans and analysts alike has already been answered: is this the year a №16 seed takes down a №1? The answer: a resounding no. That won’t come until 2030.

Let’s back up. In the history of the expanded NCAA tournament, which since 1985 has made March one of the most exciting month in sports, 16 seeds are 0–132 all-time against 1 seeds. There have been some more than close calls along the way, but no 16 seed has been able to finish the job. But we’re getting closer.

According to the Washington Post’s tournament game database, the first 13 tournaments (1985–1997) had quite a few scares for 1 seeds in the first round:

  • 1985: Michigan 59, Fairleigh Dickinson 55
  • 1986: Duke 85, Mississippi Valley State 78
  • 1989: Oklahoma 72, East Tennessee State 71
  • 1989: Illinois 77, McNeese State 71
  • 1989: Georgetown 50, Princeton 49
  • 1990: Michigan State 75, Murray State 71
  • 1996: Purdue 73, Western Carolina 71

Then things settled down. Between 1998 and 2011, the closest game was a 5-point Michigan State win over Valparaiso in 2000, actually the last year the Spartans went on to win it all. There’s a similar disparity between these time periods for 15 seeds’ triumphing over 2 seeds and 14’s over 3’s. Three 15 seeds and eleven 14 seeds pulled off the upset between 1985 and 1997, but just one 15 and five 14’s could do it between 1998 and 2011. For some reason the first round of first 13 years of the expanded tournament were much more tumultuous for 1, 2, and 3 seeds than the next 14 years. Whether there was more competition back then or the shot clock change from 45 to 35 seconds in 1993 gave top seeds more time to get back in the game, I can’t be certain.

But recently, we’ve been seeing many more 2 and 3 seed upsets as well as close 1 seed games, and it seems like a perfect storm is swelling for the first 16 seed takedown ever. Since 2012:

1 seeds:

  • 2012: 16 UNC-Asheville led 1 Syracuse with 5 minutes to play and were down just 4 with 1:30 to go when the refs failed to call a 5 second inbound violation on Syracuse. ‘Cuse made free throws and won 72–65.
  • 2013: Kansas 64, Western Kentucky 57
  • 2013: Gonzaga 64, Southern 58
  • 2014: Coastal Carolina led Virginia 35–30 at halftime but Virginia ran away with it in the second half

2 seeds:

  • 2012: Missouri lost to Norfolk State, Duke lost to Lehigh
  • 2013: Georgetown lost to FGCU
  • 2016: Michigan State lost to Middle Tennessee

3 seeds:

  • Five 3 seeds have lost first round since 2012, most notably Duke to Mercer 71–78 (2014)
Last year 15 seed Middle Tennessee State stunned 2 seed Michigan State 90–81.

As a whole, the last 20 years of the tournament have featured very few upsets/near upsets, but when they have happened, it’s mostly been the last five years. There’s no consensus as to why, but I have a few theories. The rising prevalence of one-and-dones at top schools like Duke and Kentucky has led to a couple of things:

  1. These teams increasingly rely on inexperienced freshman to play leading roles, creating problems in high pressure tournament games.
  2. Great high school recruits who aren’t quite good enough to go one-and-done aren’t playing at top programs nearly as much. They’re looking for programs that will develop then, and this trickles down to even the small conference teams that get low seeds.

Additionally, basketball has been rising in popularity over the last ten years with health concerns about football and the sinking popularity of baseball (sad!). This benefits 16-seed-worthy teams much more than Duke, Kentucky, Kansas, UNC, etc. because the top programs are going to get the best players anyway, guys who are going to play basketball over any other sport. But the small programs get a lot of the talent that could otherwise be playing football, baseball, or not at all. This causes the ability gap between top programs and small conference teams to narrow.

A late blown call cost 16 seed UNC Asheville a chance at beating 1 seed Syracuse in 2012.

Whatever the reason, low tourney seeds have given high ones more of a run for their money in recent years, and the first 16 over 1 upset is approaching. In fact, it’s pretty surprising that it hasn’t happened yet. Vegas on average gives 16 seeds about 50:1 odds (2% chance) to win, meaning all 132 1 seeds collectively have overcome about 14:1 odds (7% chance) to remain unbeaten after 33 years. By Vegas odds, it should have happened 2 or 3 times by now. Kenpom gives the massive upset even higher odds, 25:1 (96% chance) on average, with some 1 seeds receiving a less than 90% chance of winning. Ken Pomeroy does admit, however, that his algorithm gives 16 seeds too much credit.

A foolish thing here would be to fall into the gambler’s fallacy, believing that it must happen this year or next year because it’s simply due. That’s how you lose $500 playing casino blackjack because you just know your next hand has to be at least a 20 because it hasn’t happened in so long. That’s why I’m going to give this investigation a lot more statistical backing. I’ll first produce a normal distribution of the margins of victory of 1 seeds from the last 20 tournaments. Without getting too technical on you, this will allow me to observe the average victory margin, standard deviation, and probability that the victory margin is less than zero (meaning that the 1 seed loses). Here’s what I found:

  • Average margin of victory = 25.9 points
  • Standard deviation of victory margins from the mean = 12 points
  • Probability that a 16 seed had on average to upset a 1 seed = 1.56%

This would mean that it should take on average 64 games, or 16 years, for a 16 seed to topple a 1. But I wanted to do something a little more involved than this. My theory after all this that the rising rate of 2 and 3 seeds losing first round is a signal that a 1 seed will topple soon as well. Really, the 2 seeds are just as good as 1 seeds. Since 1998, 1 seeds are actually 15–17 against 2 seeds in the tournament. Similarly, I think it’s fair to assume 15 and 16 seeds aren’t too far apart in ability either. And it’s maybe not best to use all margins of victory by 1 seeds as our forecasting metric here. After all, the upset only needs to happen once. Even if three 1 seeds win their games by 40 points each, if the other goes down… well there’s the upset we’ve all been waiting for. Therefore, I’ll be using lowest margin of victory by any of the schools for each year. So because this year Gonzaga won by 20, Villanova won by 20, Kansas won by 38, and UNC won by 39, I’ll use the lowest margin of 20.

Are these big assumptions? Is this a perfect forecasting technique? Yes and probably not. But I’m predicting an event that has never happened before in a sport that thrives off of unpredictability. So I’m going for it. Piecing my two new ideas together, I’m going to create a new metric, one I’ll call a conglomerate 1 seed margin of victory (COSMOV), where I use the lowest margins of victory of 1, 2, and 3 seeds each of the last 20 years, weighted at 60%, 30%, and 10% respectively, to comprise it. It looks like this over the years:

I believe this new metric is a more accurate reflection of how likely a future 1 seed is to fall to a 16 since it acknowledges to some extent the performance of 2 and 3 seeds and puts more emphasis on the lowest rather than individual margins of victory each year. And now, the stats:

· Average COSMOV = 11.9 points

· Standard deviation of COSMOV = 5.8 points

· Probability of any 16 seed beating a 1 seed = 2%

As you can see these odds are a little bit higher now. This means that one should expect it to take 50 games to see the David over Goliath victory, or 13 years, 3 fewer than the other method predicts. So there you have it.

Now for the fun part: Which team will be the first to lose to a 16? In an extremely unscientific but reasonable way I’m going to say that it will be the team that has the greatest chance of getting 1 seeds and is most prone to large tournament upsets. That’d be Duke. Duke has gotten more 1 seeds than any other team in the last 20 years (11) and also has some very notable early-round disappointments (11 VCU in 2007, 15 Lehigh in 2012, 14 Mercer in 2014, and somewhat of a stunner this year against 7 South Carolina). I don’t know whether it’s because teams give it all they’ve got against Duke or if it’s that Coach K doesn’t do his scouting reports on big underdogs (more likely the former), but Duke teams with high hopes have a propensity to get knocked out early. As for the 16 seed I’m picking, it’s a blind guess, but I’ll go with a team who’s been there before quite a few times. So you heard it here first: In the first round of the 2030 NCAA Tournament — 1 Duke 65, 16 North Carolina A&T 66.

The North Carolina A&T Aggies, your future 16 seed insurgents.