The Upset Zone

"You are about to enter another dimension, a dimension not only of sight and sound but of mind. A journey into a wondrous land of imagination. Next stop, the Upset Zone!"* For my readers who also happen to be science fiction aficionados, the intro to this article was a PPB reproduction of the infamous intro to the Twilight Zone television series. I can promise this is not a cross-over to or application of Twilight Zone thinking to bracket picking. I just wanted something cute to lead in the article about a phenomenon that makes you feel like you've just left a bizarre and unusual place. After 2018's R64 games, you probably felt like you had just left the Twilight Zone because 2018's tournament gave us bracket history. That's right, a 16-seed has finally beaten a 1-seed. 34 years of tournaments, 135 match-ups pitting 1-seeds against 16-seeds with 8-10 near-misses among these, and the last 1v16 match-up of the 2018 R64 gave us our Haley's Comet of Cinderellas. In all honesty, I wasn't euphoric about it, even though I should have been since it has never happened before. I wasn't really shocked by it happening (the margin of victory however was a surprise) or the team to which it happened. My first gut reaction to it was the inspiration for this article: "So, now what?"

In this article, I will answer that question: "What's remaining in the Upset Zone for us bracket pickers? To begin with, I want to better understand the conditions that gave way to the first-ever 1v16 upset so that the next one is closer to a given than a guess. Second, I want to know what other historical firsts are waiting for us in the Upset Zone.

"Once Upon a Time": A case-study of the first-ever 1v16 upset

For this case-study, I will compile a list of observations (or anomalies or whatever you want to call them) in no particular order that could be used as warning signs for the next 1v16 upset.

• A 1-seed Minus One Equals: Shortly after earning the top overall seed in the 2018 tournament, UVA quickly made headlines with a key injury. Redshirt freshman and key reserve player Deandre Hunter incurred a season-ending injury. Injuries affect rotations, they affect rhythm, they affect consistency, and probably most important, they affect chemistry. If you don't believe me, check out 2012 SYR who missed starting center Fab Melo in their R64 match-up (a 1v16 near-miss) against UNCA. Injuries during the 2018 tournament derailed a few runs by F4-favored teams like PUR and GONZ. Furthermore, in my honest opinion, I think they are more devastating to system-based basketball programs like UVA (2009 PITT under Jamie Dixon who dodged a bullet against ETST).
• A Night Game with No Stars Out: 30 years from now, could you name the team that was the first 1-seed to lose to a 16-seed? If you are reading my blog, you probably will, but here's a harder one: Could you name a player on that team? I follow college basketball closer than any sport and I follow the ACC closer than any other conference, but 30 years from now, I doubt I will remember the name of any player on the 2017-18 UVA team. Why is that? They have no Star-power! 1989 Georgtown had Alonzo Mourning and 1989 OKLA had Mookie Blaylock. While I could list all of the stars from the 1-seeds that dodged the bullet, the one that bit the bullet had nobody in contention for National Awards, has nobody likely to play at the NBA level, and (maybe a good reason why they lost) had nobody that could guarantee them a bucket in desperate situations (a go-to player). The 2017 UVA team had London Parrentes, 2016 had Malcolm Brogdon, and 2015 had Justin Anderson. While 2018 UVA may have had a player named Guy, they certainly didn't have a Go-to-Guy.
• The Strength of David or the Weakness of Goliath: Was it the fearlessness of David or the over-confidence of Goliath that produced the greatest upset in history? (NOTE: I do not like applying advanced metrics in the following matter, but I occasionally do so out of intellectual curiosity.) By ranking all historical teams against all others among its seed line, 2018's 1-seeds were the 4th-weakest historically (only 2016, 2014 and 2009, in order, had weaker collections of 1-seeds). In contrast, 2018's 16-seeds were the 2nd-weakest historically (only 2016 had a weaker collection). In years where 16-seeds were historically strong (2012, 2015, and 2009, in order), no upset prevailed (although 2009 PITT-ETST and 2012 SYR-UNCA produced two near-misses). In a strange twist of irony, 2018 pitted the 4th-worst historical 1-seed (KU) against the 4th-best historical 16-seed (PENN), and while the game was close, it didn't produce our Cinderella (and it was the match-up everyone said would 'most likely' produce it). Though everyone loves to predict the next David, it seems you might be better off looking for a vulnerable Goliath.
• The Sesame Street 1-seed: "One of these things is not like the other, one of these things just doesn't belong." In my 2018 article on Metrics and Match-ups, I attempted to correlate Dean Oliver's Four Factors to a team's margin of victory and to extrapolate a team identity from the Four Factors. UVA was the only 1-seed with a defensive identity. Even a casual look at the KenPom ratings on Selection Sunday could tell you the same thing: UVA (21st AOR/1st ADR), NOVA (1st AOR/21 ADR), KU (6th AOR/46th ADR), and XAV (7th AOR/59th ADR). This may be an even more important warning sign when you consider the current meta of college basketball, which uses the rules to manufacture offense and punish defense.
• Trending Down: "One is an accident, two is a coincidence, and three is a trend." UVA played in three of the most mind-boggling games I have ever seen in college basketball: The wins at FLST and at LOU and the home loss to VT. By the non-scientific term "mind-boggling," I simply mean the final score doesn't reflect the 40 (or 45) minutes that were played, and you can see for yourself here. So, why are these three mind-boggling games important? Most teams seem to follow a rising ebb-and-flow cycle throughout the course of the season. They rise, they peak, they fall, they trough, and they repeat this pattern such that subsequent peaks and troughs are slightly higher than preceding peaks and troughs. UVA's season, on the other hand, seemed as though UVA had one really big peak (I would estimate its timing around the win vs CLEM and the win at DUKE) and one really big trough (self-explanatory). What's really important to grasp is the notion that a team can still lose while trending up (the loss to WVU) and can still win while trending down (at FLST and at LOU). Until my off-site work on cycles and trends produces fruit, this claim will have to remain conjectural.
• The End of an Era or the Beginning of Something New: 1985-1990 marked some really wild tournaments. Three of those years featured 13 total upsets, two of which produced a M-o-M rating above 18% and the third above 15% (2018 was 20.3%). Two of those years produced an 8-seed National Champion and a 6-seed National Champion (only twice since then has the National Champion been a 4-seed or worse). Most important to our current discussion, 1989 featured three near-misses in the 1v16 match-up: GTWN-PRIN, OKLA-ETST, and ILL-MCNS. Following these tumultuous six years, 1991-1996 featured relatively calm tournaments, with upsets totaling seven or less for each year and only one year featuring a M-o-M rating over 13.3%. On the other side of the coin, a 15-seed would finally knock off a 2-seed in 1991, again in 1993, and then a third time in 1997 (not a single 2-seed fell victim from 1985-1990 when tournaments were wilder). I point all of this out because it seems to fit our current era in tournament history. From 2010-2018, tournament insanity has been the norm, maybe worse than 1985-1990. In these nine years, only two tournaments have registered M-o-M ratings lower than 17% and only two tournaments have totaled single-digit upset totals. Maybe the era itself contributed to the first-ever 16-seed upset?!?!? As a friend of mine described it, "We were long overdue!" If the near-misses of 1989 and the era of craziness from 1985-1990 led to the results of 1991-1996, then how does 2018 UMBC-UVA and the era of insanity from 2010-2018 translate going forward?

I want to address one important detail about this list. Each of these criteria was present in one or more of the near-misses. Only all six were present in the actual upset. It is why this list doesn't appear to be exhaustive. One I really wanted to key on was UVA's deliberate tempo-killing offense, but in the end, I chose to consolidate my own lengthy list of factors with the ones I believe to be significant for future 1v16 upsets. If you feel that I've over-looked one, feel free to supply and support it in the comments section.

"What's in the Box": The remaining UPMs

Let's look at the damage that the 2018 tournament did to the Upset and UPM totals, and then we will take a look into the box.

If you read the update to the Historical Approach to Upsets, this table should look familiar. Here are the key takeaways:

• I've spent the whole article thus far talking about one specific R64 match-up, so the only other R64 UPM worth mentioning is the 5v12. Coming into 2018, the gap between the 5v12 and 6v11 UPMs was two, but with the 5-seeds holding serve (like I predicted they could) and the 6-seeds splitting the pair, the gap between them expanded to four.
• In the R32, the 1v9 UPM gained some ground against the 1v8 with the FLST upset over XAV (I failed to predict that). The 2v7 also gained some ground against the 2v10, and it should considering our postulate on seed differentials. The 3v11 went 0-2 against the top-seed, which even surprised me considering how strong the 3-seeds looked on paper.
• In the S16, 2018 gave us another first, but unfortunately, I do have to qualify it. We had our first ever 4v9 upset with FLST defeating GONZ, and now that you know the individual game responsible for it, you should know the qualification I'm about to make. GONZ lost key starter Killian Tillie, whose ability to stretch the floor with his perimeter shot from the power forward position would have been invaluable to GONZ against the height and length of FLST. Oh yeah, the 5v9 moved from 1 out of 2 (50%) to 2 out of 3 (67%) and the 7v11 is still batting 100% after its fourth try in 2018.
• The E8 only had one UPM: A 3v8. No upset here.
• The F4 also had one UPM: A 3v11. No upset here either.

The Upset Zone: So what upsets do we have remaining?

• Let's start with the obvious ones: 16v8 and 16v9. Since we got our first 1v16 upset, it is only natural that we start looking for the 16-seed to pull off the upset in the next round. We could be here for a while though. It took seven 2v15 upsets before a 15-seed won in the R32, and we do know that was largely due to erroneous seeding by the committee. We probably should have had another 15-seed upset in the R32 in 2015 when MTSU was erroneously given a 15-seed. Likewise, we also could have got our first 16v9 upset in 2018 considering the injuries facing 9-seed KNST going into that game (KNST only won 50-43). Until then, we'll be waiting for our first 16v9 upset and our first 16v8 UPM. First, we have to find our 2nd 16v1 upset.
• Since I've already mentioned its counterpart that has happened, now is the perfect time to mention the 15v10 upset. With five attempts at an upset (two more than the 15v7 UPM), the 10-seeds have held serve every time. One of them will crack, but the real challenge is finding the 2v15 upset first.
• In a Twilight Zone twisting fashion, let's look at the UPMs that need some mean-reversion.
• S16 7v11: I think this one is the most out-of-hand. Four attempts resulting in four upsets seems more bizarre than if 1-seeds had ended up being 136-0.
• R32 2v10, 3v11 and 4v12: Since 2v7 with its seed differential of 5 has a 30.49% upset rate on 82 attempts and these three UPMs and their seed differential of 8 each have higher rates on half of the attempts, these three may need to come back down to earth. I hope the 2v7 UPM isn't trying to make-up ground with 5 upsets in their last 9 tries.
• E8 1v11: How on earth are 11-seeds better against 1-seeds than 6-seeds? For perspective, all have occurred in relatively insane years (1986, 2006, and 2011). Either way, this upset rate seems inflated. In fact, the E8 2v8 rate could use some deflation as well.
• I should also take this opportunity to point out the UPMs that haven't happened yet (in the table above, they will have "N/A" listed in the "18U%" column). The S16 UPMs include the 2v14, 15v6, and the 15v11, and the E8 UPMs include the 4v11 and 5v11. What a wild tournament it will be if these teams are meeting in these rounds (even assuming all other match-ups feature the best possible seeds).
• I thought I would save the best for last. Since the 1v16 used to be the Granddaddy of All Upsets before 2018 happened, we now need a new one. After 19 attempts and no dice, our new creme de la creme is the S16 UPM 1v12. Very few plausible match-ups feature a seed differential of 11 (the only other one being 3v14 in the R64). After knocking off the perennial victim 5-seed and most likely going through the 4-seed in order to face the 1-seed, the 12-seed has stolen the nation's attention and most likely the 1-seed's attention too. The only time a 12-seed has even been to the E8 was through an 8-seed. There would be no better way to announce the 12-seed's arrival into the E8 than by knocking off a 1-seed. The best part of this upset if it ever happens: I'll get to do another Upset Zone article.
• I promise that I'm done after this one. It took 7 tournaments before we saw our first 15-seed knock off a 2-seed. It took another 21 tournaments before we saw two 15-seeds knock off two 2-seeds in the same tournament. How long before we see 16-seeds take 50% of their R64 UPM? If it took 15-seeds three times as long from their first-ever upset until their R64-split, then it means we should expect the 16-seeds to accomplish their R64-split in 102 years. I can only hope PPB will still be around to analyze this feat because I know I won't be.

I hope you enjoyed this article as much as I did. I would like to point out that I won't make the entire 2019 PPB agenda about upsets like I did in 2018. I don't regret 2018's focus on upsets because it helped me assess the 2018 tournament better. Not to mention, the penultimate upset happened in 2018, so the focus on upsets was justified. If 2018's focus on upsets helped any of my readers pick it in their brackets, then it was definitely worth it. However, 2019 will take a more balanced approach, but the goal will still be the same. As always, thanks for reading my work, and the next article should be published around the beginning of December.

Works Cited
*The quote reproduced is given credit and attribution to The Twilight Zone®. Retrieved from source: https://en.wikiquote.org/wiki/The_Twilight_Zone_(1959_TV_series).