O n a Wednesday night in March during my final semester as an undergrad, I pulled an all-nighter laboring over something that had no impact on any of my final grades whatsoever. March Madness was upon me and I’d convinced myself that if I combed through enough data (neutral site win percentages, shots per game, time of possession, etc), I could develop an algorithm that would help me create the perfect NCAA Tournament bracket. In my mind, I was like Russell Crowe in A Beautiful Mind, meticulously cracking codes that would ultimately reveal the secret to correctly picking upsets. To my roommates, I was more like Charlie Kelly in It’s Always Sunny in Philadelphia, maniacally blurting out incoherent exclamations that sounded more like a mailroom conspiracy theory than anything remotely resembling sound logic.
I stayed up so late working my way through each of the tournament’s 63 matchups that when I finally submitted my picks and retired for the night (morning), I slept through the first three games of the tournament in their entirety. Much to my bewilderment, I awoke to the news that each of my first three picks had been wrong. Less than three hours into a tournament that spans three weekends, my attempt at the immaculate bracket had already fallen flat.
It took me a while to get over that. I certainly don’t put anywhere near the effort into tournament picks now that I did that year, though to be fair, that’s a pretty high bar to live up to. I thought about my quixotic quest for bracket immortality the other day, though, as this year’s field of 68 was revealed. More generally, I thought about the way that we, the bracket pool contestants, approach filling out these damned things, and wound up stuck on something: Why do we even pick upsets? What is it that drives us to go against all seeming statistical logic year after year to try and achieve something that’s next to impossible? As it turns out, it’s a lot more logical than it seems.
Picking upsets in hopes of creating the perfect bracket is a fundamentally risky endeavor. Statistically, it’s a fool’s errand. The odds of creating a flawless bracket are 1 in 9,223,372,063,854,775,880 and yet each year millions of people try and do just that. Knowing how unlikely it is to compile a string of 63 correct picks, you wouldn’t fault someone for throwing their arms up in frustration and simply picking the favored team in each game. On face, at least, that seems like the rational thing to do. But humans aren’t always rational creatures.
W e take risks because inherent in them is a reward. When we’re confronted with a choice whose outcome is uncertain, we do some quick mental accounting and decide whether the potential gain outweighs the potential loss. This is known as prospect theory, a concept proposed by Daniel Kahneman and Amos Tversky, two of the world’s leading behavioral economists. The theory holds that as humans, we’re hyper-sensitive to the prospect of losing out on something and consequently tend to opt for outcomes that present the lowest chance of that loss coming to fruition.
If–for example–I were to present you with the choice of either taking $800 guaranteed right now or risking making a bet that had an 85% chance of winning you $1000 (and a 15% chance of winning you nothing), you’d probably take the sure thing: the $800. That’s because by nature, we’re risk-averse. In general, we prefer a sure outcome over a gamble, even when that gamble can potentially be worth more than the sure thing.
By that logic, it doesn’t really make much sense that we pick upsets–often a lot of them–each year. After all, higher-seeded teams are favored for a reason and if we take the tournament selection committee’s seeding decisions as a representation of a team’s chances to win a game as fairly accurate predictors, it’s in our interest to steer clear of the underdogs. But the NCAA tournament doesn’t work like that. It functions in a way that encourages us to do the opposite of what seems to make sense.
The round of 64 features 32 different games for which people filling out brackets need to make picks. If we think of each game’s result as an event that contributes to an overall point system–like you’d have in an office pool–the obvious strategy (assuming you’re trying to win) would be to accumulate as many points as possible. Picking the favorites in every game, in theory, gives you the best chance to do that. But there are a few problems with that strategy, most notably the fact that first round upsets do happen.
From 1985–when the tournament expanded the field of teams to 64–through to last year’s tournament, there have been a total 156 upsets in which teams seeded 11 or higher knocked off their first-round opponent. That’s nearly five upsets per year just in the first round. Last year’s results alone featured two 11-seeds advancing (Loyola-Chicago; Syracuse), two 13-seeds advancing (Marshall; Buffalo), and, for the first time in tournament history, a 16-seed knocking off a 1-seed (UMBC).
For fans filling out their brackets, those kinds of results are hard to ignore. What, on paper, is supposed to be pretty close to a sure thing is, in reality, far from it. The George Masons, Florida Gulf Coasts, and Norfolk States of the world aren’t just theoretical outcomes, they’re distinct possibilities, if not probabilities. When we know that a round of 64 upset is bound to pop up, it completely inverts our decision-making process.
Returning to the example above, what if I now told you that you had to choose between a guaranteed loss of $800 or a bet that offered an 85% chance of you losing $1000 but a 15% chance of you losing nothing? This time, you’re likely to take a risk instead of accepting the flat $800 loss. This is risk-seeking behavior, another facet of prospect theory. Those who engage in this kind of behavior see the prospect of a guaranteed loss as less-desirable than the possibility of losing even more in pursuit of an outcome that could mean that they lose nothing at all. Let’s put that in bracket terms.
I’m making my round of 64 picks with the knowledge that upsets are pretty much a sure thing. I have two options:
1) I can stick with my initial strategy of only picking the higher-seeded teams. I’ll probably take a few L’s in the process and my dream of a perfect bracket will almost certainly be shattered, but I can at least limit the overall damage to my bracket, assuming the teams that cause the upsets don’t make it too far in the tournament.
2) I can try to avoid potential losses altogether by opting for the lower-probability option of picking upsets. Doing this gives me a chance to avoid bracket damage completely–keeping the window open for me to realize my dream of pick ‘em perfection–but it also exposes me to even more risk than simply going with the favored teams if I pick the wrong upsets.
Ultimately, the prospect of ending up with a bracket that’s almost certainly no longer in-tact when we feel there’s something we could have done about it–however small the probability of that “something” actually working may be–is enough to make us pencil in North Dakota State over Duke and a second round bow-out for UNC.
As with so many other facets of our lives, we’re slaves to decision-making that doesn’t always make the most sense. In their exploration of the process through which humans make decisions, Kahneman and Tversky underscored the fact that we’re twice as sensitive to the prospect of losing something as we are to gaining something. Put another way, we’re unlikely to agree to flip a coin if doing so means we could lose $10 unless a successful coin flip results in us winning more than $20. That same sensitivity to losing has me making picks I know I shouldn’t, seeking risk wherever I can find it, and hoping for outcomes I know won’t happen. It’s Senior year all over again and I just can’t help myself. Let’s go, North Dakota State!