Whoa

This is fascinating.  Over on Cognitive Daily, they posted the results of a quick experiment in which readers were asked to select a number, any random number they wanted, between one and twenty.  Here were the results:



For whatever reason, an extremely high percentage of the 347 respondents picked 17 (dark blue bars).  By comparison, 347 numbers randomly generated by a computer (light blue bars) were much more evenly distributed.  Significant analysis follows, and several objections to the conditions of the experiment are posted in the comments. 

Mostly I think it shows that humans and computers have different ideas about what "random" really means.  Many readers probably felt like 17 was the least common number in daily life (although if you're a cribbage player, you'd argue that it's 19) so therefore it seemed the most random.  Whereas you can see that 5, 10, 15,  and 20 all got a low response.  A machine wouldn't differentiate the same way.

So speaking of random number generation, every year I participate in a college bowl game-picking pool organized by a friend of mine.  Essentially, it's just a contest between about twenty people to try to pick the most bowl games correctly.  I only vaguely follow college football so I'm out of my depth to begin with, although the pool is comprised of a range of people, from those obsessed with college football, down to people who are completely ignorant.  I'd guess I'm in the 20th percentile or so of participants.  Meaning I have more base subject knowledge (on college football) than about 1 in 5 others.  But I think anyone who's participated in a NCAA tournament pool knows that knowledge can be a dangerous thing, and the person who has followed the game closest throughout the season not only doesn't have an advantage, they probably, in fact, have a disadvantage. 

The point is, I make all of my picks via random number generation.  I've tried different methods in the four years I've played, all lacking scientific validity.  Here's a summary of my methodology:

Year 1: Random number generation by MS Excel
Year 2: Combination of Excel and polling my roommate (whom I perceived as lucky) for random numbers
Year 3: Asking friends to gather numbers from random.org, which I compiled and gleaned results
Year 4: Random number generation by Google spreadsheet

This last year I finished in 2nd place in the pool.  Second place!  Picking entirely randomly, going only by meaningless numbers generated by Google spreadsheet software, I out-picked nearly the entire field.  Year 1, the other year with no human intervention, was similarly successful--in the top three I think.  The worst year was Year 3, which required the most human participation.  Even though the numbers were random, I asked people to gather them.  I finished last, easily.  Year 2 was somewhere in the middle.

What I'm saying is, if you're in an NCAA tournament pool with me this year, beware Google spreadsheet!

Posted at 09:56AM Feb 09, 2007 by WILSON, JOSHUA in Thought experiments  |