Always pick the higher seed. Never pick an upset.
The behavioral phenomena is called probability matching, explains Matthew Hutson.
Let’s say you’re drawing balls from a large box that contains 25 red balls and 75 blue balls. Many people tasked with predicting draws will predict red 25 percent of the time and blue 75 percent of the time. They match their guesses to the probabilities of the outcomes, in this case yielding a 62.5 percent success rate. But if they just guessed blue on each draw, they’d be right, on average, 75 percent of the time.
Similarly, people know there will be a certain number of upsets in each NCAA tournament, and therefore betting on a Cinderella-free tourney seems silly. The only logical thing to do, they conclude, is to figure out when those upsets will take place. The drawback, of course, is that by shooting for perfection, they end up handicapping themselves. In McCrea and Hirt’s research using real NCAA data, for example, people would have been much better off just sticking to the seedings.
One thing is for sure. You won’t have much fun filling out your bracket with this strategy.