Output Stream of Consciousness

Link: http://www.amazon.ca/exec/obidos/ASIN/0002007916/thecerealkill-20

This is a very entertaining, easy-to-digest (don't eat it) book all about, as the subtitle of the book says, "The Curious World of Probabilities".

The reading is brisk, and he goes over a number of pretty practical subjects. The Law of Large Numbers figures large in this tome. That is, the more times a probability event happens, the closer it gets to the calculated odds. So, you may flip three heads in a row, but flip the coin 50 times, and you'll get 23-27 heads almost guaranteed...

There's a goodly amount, not surprisingly, about gambling in here. You might be surprised to find out that the house wins, but not a whole lot more than 50% of the time. They use the Law of Large Numbers to good effect. You might be surprised that the least bad chance of loss in a casino (by and large) is in craps, with an average 14 cents loss per ten dollar bet. Of course, you don't literally lose 14.1 cents on a ten dollar bet, but if you kept playing, and maybe spent $10,000 on craps, you'd probably lose about $140. Repeated over and over again, casinos can live large. Keno is probably the worst.

There's much ado about winning lotteries (many of which really have odds such that you stand more chance of dying in the next 30 minutes than you do of winning), risk of dying of terrorist attacks even in a seemingly dangerous place like Israel (wildly overestimated).

There are good sections in the book about political misuse of statistics (murder rates have gone down steadily over the last decade in the US - but that's not the statistic they will use), medical statistics for new products (medical data can be good, but publishing bias and other biases can totally and dangerously skew the implications), the upside of randomness (privacy, games, simulations), and debunking the myth of "if you're past 40, you have more chance of being killed by a terrorist than of being married" (besides the fact that the data is from 1970 :) ).

The chapter on the Monty Hall problem was surprising. If you pick a door, Monty Hall showed you a door your prize was not behind, and then gave you the chance to change your choice. According to Marilyn vos Savant, you should change your choice to the other uncovered door. You will win the prize two out of three times.

The chapter on utility functions was a bit fluffy for my tastes. Utility functions can be used to 'quantify' happiness or whatever else have you - boss shouting at you, -200, lunch at The Keg, +60, etc. No good guidelines as to scale or what have you. The coverage of probability and randomness in biology and genetics was also exceedingly light - I realize he might not want to go out of his domain here, but there's much material and expertise elsewhere he could have drawn upon. I was a little disappointed at his simplification of evolution down to the Law of Large Numbers:

The Law of Large Numbers ensures that in the long run, each generation of this life form will, on average, be a little more "human" than the previous generation.

More context could have made it clear what he was getting at.

He's got illustrative 'stories' for a lot of the main points he is making, and they're pretty good at illuminating the practical side of things, and there are a few film noir stories of using probability in a detective capacity which are just great :)

A nice, distracting read :)