Want bracket success? Don't outsmart yourself

Mon, Mar 15

Put down the agate-type team capsules. Stop studying 3-point shooting percentages. If you want the best chance at winning your NCAA tournament office pool -- held purely for nonmonetary, recreational purposes, of course -- only one bracket-picking method makes sense.

Go chalk.

Selecting upsets is fun; predicting the giant killer no one else saw coming (such as Santa Clara over Arizona in 1993) can secure intracubicle bragging rights for life. As an overall strategy, however, picking upsets is not only dumb but the uniquely human kind of dumb that comes from outsmarting ourselves. So says a recently released study from Indiana University and the University of Wyoming that finds that:

1. People pick too many upsets, as their average predictive performance fares worse than the tournament seeding and no better than chance;

2. People can't help but pick too many upsets, via a phenomenon known as probability matching.

"The bottom line is that we think we can do better than the seedings," says Ed Hirt, one of the study's authors and a professor in Indiana's department of psychological and brain sciences. "We know more and can figure out who is hot, who is overconfident or a choker in the tournament. But all the information we have is likely already accounted for in the seedings anyway, so our illusory belief that we can do better actually hurts our performance."

First, the numbers: Looking at the average performance of individuals in the 2004 and 2005 ESPN.com Tournament Challenge, the study found that participants were 75.2 percent correct in '04 (72.9 percent in '05), while going chalk would have resulted in being 87.5 percent correct in '04 (75.0 percent in '05).

If NCAA tournament upsets are essentially random, then why the insistence on making lousy picks? Hirt attributes much of it to probability matching, which in the context of the NCAA tournament basically means that individuals are aware that certain upsets seem to happen regularly and ill-advisedly make picks accordingly. Take the 5-12 games: Most people feel compelled to pick one upset -- it happens every year! -- even though they'd be statistically better off to select the four favorites.

"The best analogy I can make is a student who has to guess at the end of a multiple-choice test where they weren't able to get to all the items," Hirt says. "Your best bet is to guess the same response for all the rest, so that you get at least some of them right.

"However, people may think that the teacher tends to have certain probabilities of answers being correct. You think your teacher has a 20 percent A, 20 percent B, 40 percent C and 20 percent D, so if you have 10 blank ones, you guess two A, two B, four C, and two D. By doing so, you may get none of them right -- but if you put C all the time, you would get four correct. That's how I think of it."

And that, in turn, brings up another good reason to go chalk: It's less of a headache.