The Bracketology of Crowds, Part V: Final Four and Championship Results

You will recall from my last post that so far the Crowd has done no worse than 75% in a given round of the tournament.  For the Final Four and Championship game, the Crowd made a clean sweep, 100% in each round (not so astounding when you're only talking about 3 games, but still).

The Crowd successfully picked 49 of the 63 games, an impressive 77.8%.  That put my bracket well out in the lead (in terms of accuracy) in my own pool of 5 contestants (my closest competitor had 40 correct picks).  Due to the vagaries of our scoring system, though, I finished 4 points out of first place.  Go figure.

Like any good economist I was a little skeptical of comparing myself to so small a pool, since small sample sizes are notorious for giving spurious results.  So I checked myself against a larger pool that a coworker had joined, one with 45 participants.  The competition is fiercer, but the Crowd still performs astoundingly well.  The most accurate brackets in this larger pool picked 49 games correctly, which leaves the Crowd tied for first place and ahead of almost 98% of the participants.  The highest score was 298 points, which edged out the Crowd by a single point!  Apparently it's my fate to finish second despite the highest level of accuracy.  Still, the Crowd's score beat out about 95.5% of the participants.

I'll spend another post reviewing some of the weaknesses of my approach to 2007 March Madness, but it's pretty clear that the Wisdom of Crowds performed quite well in this context:  finding a good system to aggregate everyone's knowledge outperformed the majority of intelligent individuals. 

Comments

[Todd Easton]

The crowd beat me.

One characteristic of most pools of this nature is that only first-place wins (or maybe 2nd & 3rd). This causes most entrants to increase their tolerance of risk. The people who win large pools (generally) have to predict a few upsets to differentiate from the "crowd" pics. There is a natural incentive for people to make irrational picks.

To your point, however, the risk/reward scale in business doesn't only pay off for the top performer. As your experiment demonstrates, in business, one can acheive more-superior results by building on the aggregate knowledge of others.