The Bracketology of Crowds, Part II: 1st Round Report

Round 1 is over; here's the rundown of how the Crowd is performing in my bracket pool.

The crowd correctly picked 25 of 32 games (a little better than 78%).  In my bracket, that has me tied for first in accuracy.  Unfortunately, the crowd also shunned upsets for the most part, and since my pool includes the team seed in the scoring process, this has hurt my ranking.  I'm currently in third place (absolute middle of the pack), two points out of second place, and 21 points out of first.

But here's where things get interesting.  So far, all of the crowds mistakes are what I call "shallow."  That is, none of the teams the crowd incorrectly picked to win a Round 1 game was also picked to win Round 2.  This means it could potentially pick the remaining brackets with 100% accuracy (unlikely, but possible).  Contrast this to my opponents.  A certain director of MBMI (currently ranked 2nd) picked Creighton to make the Elite 8 and Villanova to make the Final Four.  That's likely to kill his score in the coming rounds.  A second opponent predicted Duke to make the Final Four, which will also be costly in the coming rounds.  A third has so many deep losses in the West that that bracket is essentially dead for the remainder of the tournament.  In short, while the crowd's score is currently fair-to-middling, it's long-run prospects look pretty good.

So far, the crowd's greatest virtue appears to be its restraint.  It's easy for an individual to get excited about a team like Duke, which has a long history in the Tournament.  In fact, the crowd wrongly picked Duke to beat VCU.  But where an individual might easily go on to assume Duke could come off that victory to beat 3-seeded Pittsburgh, it's hard to get the crowd that riled up about a 6-seed.

More tomorrow on the success of this Knowledge Process.  Until then, let's all toast the downfall of the overrated (I'm told) Blue Devils to the mighty mouses of Virginia Commonwealth.

Comments

[Eliot]

It says, in Wisdom of Crowds, that crowds have special predictive power only when the individual judgments are rendered independently, without the individual predictors conferring with each other. But the situation you describe violates this premise: While people are picking their brackets, they have access to what other people have already predicted; and their own judgment is shaped by it. So their judgments are no longer independent.

[Abel Winn]

Elliot,

You bring up an interesting point. Not every user of Sportsline.com is an independent observation; some users are probably relying on the opinions of others in making their predictions (though most likely not to the extent that I am).

However, with the very large number of users, it's likely that there is also a large number of independent predictions. So it's unlikely that all of the crowd's wisdom is being warped or dissipated.

Keep in mind that similar stock markets and betting markets also give participants access to the predictions of others. Yet these markets seem to do an excellent job at aggregating knowledge. Similarly, the Crowd is performing pretty respectably in the Tournament (this year, at least).

Admittedly, though, there are some flaws to my method, and I'll address them in a post after the tournament is complete.