Tag Archives: brackets

Brackets, Preferences, and the Limits of Data

As you may have heard, it’s March Madness time. If I had to guess, I’d wager that more people make specific, empirically testable predictions this week than any other week of the year. They may be derived without regard to the quality of the teams (the mascot bracket, e.g.), or they might be fairly advanced projections based on as much relevant data as are easily available (Nate Silver’s bracket, for one), but either way we’re talking about probably billions of predictions. (At 63 picks per bracket, we “only” need about 16 million brackets to get to a billion picks, and that doesn’t count all the gambling.)

What compels people to do all of this? Some people do it to win money; if you’re in a small pool, it’s actually feasible that you could win a little scratch. Other people do it because it’s part of their job (Nate Silver, again), or because there might be additional extrinsic benefits (I’d throw the President in that category). This is really a trick question, though: people do it to have fun. More precisely, and to borrow the language of introductory economics, they maximize utility.

The intuitive definition of utility can be viewed as pretty circular (it both explains and is defined by people’s decisions), but it’s useful as a way of encapsulating the notion that people do things for reasons that can’t really be quantified. The notion of unquantifiability, especially unquantifiable preferences, is something people sometimes overlook when discussing the best uses of data. Yelp can tell you which restaurant has the best ratings, but if you hate the food the rating doesn’t do you much good.*

One of the things I don’t like about the proliferation of places letting you simulate the bracket and encouraging you to use that analysis is that it disregards utility. They presume that your interests are either to get the most games correct or (for some of the more sophisticated ones) to win your pool. What that’s missing is that some of us have strongly ingrained preferences that dictate our utility, and that that’s okay. My ideal, when selecting a bracket, is to make it so I have as high a probability as possible of rooting for the winner of a game.

For instance, I don’t think I’ve picked Duke to make it past the Sweet Sixteen in the last 10 or more years. If they get upset before then, my joy in seeing them lose well outweighs the damage to my bracket, especially since most people will have them advancing farther than I do. On the other hand, if I pick them to lose in the first round**, it will just make the sting worse when they win. I’m hedging my emotions, pure and simple.***

This is an extreme example of my rule of thumb when picking teams that I have strong preferences for, which is to have teams I really like/dislike go one round more/less than I would predict to be likely. This reduces the probability that my heart will be abandoned by my bracket. As a pretty passive NCAA fan, I don’t apply this to too many teams besides Duke (and occasionally Illinois, where I’m from) on an annual basis, but I will happily use it with a specific player (Aaron Craft, on the negative side) or team (Wichita State, on the positive side) that is temporarily more charming or loathsome than normal. (This general approach applies to fantasy, as well: I’ve played in a half dozen or so fantasy football leagues over the years, and I’ve yet to have a Packer on my team.)

However, with the way the bracket is structured, this doesn’t necessarily torpedo your chances. Duke has a reasonable shot of doing well, and it’s not super likely that a 12th seeded midmajor is going to make a run, but my preferred scenarios are not so unlikely that they’re not worth submitting to whichever bracket challenge I’m participating in. This lengthens how long my bracket will be viable enough that I’ll still care about it and thus increase the amount of time I will enjoy watching the tournament. (At least, I tell myself that. My picks have crashed and burned in the Sweet Sixteen the last couple of years.)

Another wrinkle to this, of course, is that for games I have little or no prior preference in, simply making the pick makes me root for the team I selected. If it’s, say, Washington against Nebraska, I will happily pick the team in the bracket I think is more likely to win and then pull hard for the team. (I’m not immune to wanting my predictions to be valid.) So, the weaker my preferences are, the more I hew toward the pure prediction strategy. Is this capricious? Maybe, but so is sport in general.

I try not to be too normative in my assessments of sports fandom (though I’m skeptical of people who have multiple highly differing brackets), and if your competitive impulses overwhelm your disdain for Duke, that’s just fine. But if you’re like me, pick based on utility. By definition, it’ll be more fun.

* To be fair, my restaurant preferences aren’t unquantifiable, and the same is true for many other tastes. My point is that following everyone else’s numbers won’t necessarily yield you the best strategy for you.

** Meaning the round of 64. I’m not happy with the NCAA for making the decision that led to this footnote.

*** Incidentally, this is one reason I’m a poor poker player. I don’t enjoy playing in the optimal manner enough to actually do it. Thankfully, I recognize this well enough to not play for real stakes, which amusingly makes me play even less optimally from a winnings perspective.