Statistics and the manipulation of them have interested me since I saw Freakonomics writer Steven Levitt speak about street gangs and how they're like McDonalds.
The theme of superiority of super crunchers versus domain experts seems to take a bit of a somewhat ironic turn in one of the examples. On page 196, Ayres begins a discussion on the effectiveness of Las Vegas sports bookies based on work by Justin Wolfers. In a nutshell, when the point spread was below 12 points, the chance of a team beating the spread seems to fall almost perfectly into the predicted bell curve.
However, if the point spread was more than 12, the chance a team would beat the spread fell slightly left of the mean. Justin Wolfers hypothesized that there is point shaving by bribed players. He investigated and found that the point spread was right on track until five minutes before the end of games and felt this could supports the hypothesis. With a large padding with only five minutes to go, a bribed player could shave points without risking a loss for the team.