Four Ways to the Final Four

I'd like to spend this entry discussing what I view as four prevalent strategies that people use when filling out brackets for the NCAA Tournament, particularly for involvement in a pool in which points are awarded for successfully picking all the winners in the tournament, without special bonuses for picking upsets. Which category do you fall into? Are there other significant categories that I've missed¹?

1. Pick each game backed on the matchup. Ths is the straightforward strategy employed by people who feel they know enough about the teams in the tournament to predict how each game will go. They look at their brackets and say to themselves, "Hmm; is there any chance that Albany's 7' 1" big man can control the game against UConn? No? Then I'll go with the Huskies." After repeating such a dialog for the 32 first-round games, they then move on to the second round, and examine the matchups that have been formed by their first-round predictions. This continues all the way through to the championship game, at which point they apply a point-by-point to the Duke–UConn matchup and emerge with their pick for the tournament winner.

2. Pick each spot on the brackets based on aggregate matchups. With this strategy, a person chooses the team to put in each bank spot in the bracket based on which of the eligible teams is most likely to defeat the combination of other teams that they might face to reach that point in the tournament. As with the first approach this strategy relies on relatively detailed knowledge of the teams in the tournament and reasonable analytic abilities to determine which teams benefit most from their aggregate potential matchups.

3. Ask the experts. A person using this strategy spends the days between Selection Sunday and the start of the Tournament listening and reading various experts' opinions on upset specials, Cinderella teams, sleeper picks, and favorites. When he sits down to fill out his brackets, select nuggets of expert prognostication permeate his thoughts and drive picks of certain lower seeds winning first-round games and even reaching the Sweet 16, Final Four, and beyond.
4. The game theoretic approach². The previous three strategies all use different tactics to do their best to predict which team is most likely to end up in each spot on the brackets. Contrastingly, this strategy couldn't care less about the relative likelihoods of different teams winning particiular games or the entire tournament. Instead, this approach attempts to optimize the bracket's chance of winning an NCAA pool. There are a few components to this:
   • Don't choose heavy favorites. Suppose that when filling out your bracket, you believed that there was a 50% chance that Duke would win the tournament (ridiculously high, but just suppose). Any of the other three strategies would most likely result in your bracket predicting that Duke wins the tournament. However, there's a good chance that, say, half of the people in your pool are also going to choose Duke. If Duke wins (50% of the time), the winner will come down to whoever picked the best in the other games throughout the tournament. Without (or even with) any sort of expert status in the realm of college basketball, I feel that success picking the early round games that end up deciding the "tiebreak" between people who chose Duke is more or less arbitrary. So, in a twenty person pool, your chance of taking first place is only 5%. In a larger pool, with more people picking Duke, the chance is even lower.
   • Aim for uniqueness. I suppose that this strategy is as much a social strategy as a game-theoretic strategy. For certain, it would break down if competing against other people using the same strategy. However, when in a pool with 'normal' people who are sure to be exposed to media reports of how Tennessee doesn't deserve a two seed or how Memphis is by far the worst number one seed, it's not too difficult to figure out which of the better teams (better seeds) in the tournament are unlikely to be picked by many people. Memphis might only have an 8% chance of winning the tournament (true odds were closer to 6.6% at tournament start according to Tradesports), yet if I'm the only person in the pool who picks them, I'm almost guaranteed to win the pool if Memphis wins the championship. Basically, this strategy takes advantage of the fact that a (relatively) small group of pool participants is often likely to overrepresent the favorites in their picks compared to the true odds.
   • Minimize what needs to go right. This point is really subsumed by the others, but I feel it's worth mentioning anyway. There's no point to making an entry that requires three Cinderella teams to reach the Final Four when you can just as easily pick a single Memphis-esque team to win the entire tournament. Why give yourself three unlikelihoods that need to go right when one will do?

The latter two strategies, of course, are not prescriptive for every spot in the brackets. Personally, I like to fallback on my irrational dislike of Big Ten and Big East schools combined with a soft spot for the ACC and Pac 10 in guiding the rest of my picks. Other people like to use uniform colors, mascots, or hometown weather conditions to pick these less meaningful games—whatever floats your boat...

¹Raymond Chen's method doesn't count due to not being particularly "prevalent," though it is rather entertaining.

² I've never actually been trained in game theory, or even read anything about it, so perhaps this strategy is poorly named. I'm pretty sure I've laid eyes at some point in my life on people who have taken game theory classes, though.