Wednesday, March 15, 2006

The Wisdom of Crowds 1

I decided to see how far a person could go in the NCAA brackets. I made my predictions having watched zero games this year and checking three web sources. However I'd be willing to bet that my predictions beat around half of the experts if not more. I made three brackets. One bracket was straight favorites in seeding. This is testing the knowledge of a group of experts (those who make the seeding). The next bracket was based on RPI which adjusts a teams winning percentage to strength of schedule. This is betting based on past performance. Then my last bracket was based on the line at this trading market this is the widom of crowds. After each round I'll compare the brackets to those of various experts who posted their brackets. An interesting thing about these brackets is when telling others my idea they revealed their misunderstanding of probability and game theory. Their view was that there are going to be upsets so therefore you can't pick only favorites. The problem with that is that of course if you had to pick a winning percentage for your picks it wouldn't be perfect but you don't know who'll lose. As an example say you made 100 picks each with a 60% chance in your favor. If you pick all the favorites you can expect to get 60 right. If you pick only 60 favorites you expect to get 52 right (.6 times 60=36 plus .4 times 40=52) You have lost 8 points because you assumed the games were dependent variables not independent variables. This seems to be how people act as a Yale study shows. They had a T-Square maze where there was no pattern except for a 60-40 split of right to left. The rat eventually hit its maximum potential at 60%. However the students were stuck at 52%. There are two possible reasons they got to that exact percentage. One is that they found patterns and 60% of the time the patterns pointed to the right side. The other possibility is that they subconsciously realized the odds but didn't realize how to act on them efficiently they only knew to bet on one 60% of the time.


At 11:32 PM, Anonymous rachelsoloveichik said...

Most likely, a couple of students figured out the split, and always bet on the right. The rest got 50/50.


Post a Comment

<< Home