Monday, June 27, 2005

Full House

I just finished "Full House" by Stephen Jay Gould.

I'm a fan of Gould's writing - he manages to bring together very different subjects and weave them in a very interesting tale. Most of his books (that I've read) are collections of his writings in Nature. He wrote some 200 essays.

This book, however, is not a collection, but instead an explanation of a "eureka" moment he had in looking at extremes in surviving cancer, evolutionary trends, and the disappearance of batting .400 in the major leagues. You might think those three things have nothing to do with each other, but Gould shows you how they are all a product of focusing on extremes and using them to show trends that do not exist.

The baseball example (I know, baseball is boring, don't get me started) is interesting because the sport really hasn't changed in 100 years. Plus, baseball is all about statistics. In the 20's and 30's, batting .400 was not uncommon, but since then, nobody has batted over .400, rarely even approaching that goal. People often attribute it to the idea that batters just haven't gotten any better - whereas the pitchers/fielders/everyone else has gotten better. Gould lays out an argument that shows the disappearance of .400 hitters is actually a result of the entire sport of baseball getting better. Batters have averaged .260 (more or less) for 100 years, and the bell-curve distribution of batting averages has just gotten tighter (the standard of deviation had gotten smaller). As a result, both the "best" and the "worst" are closer together than ever. The same thing can be shown for fielding, pitching, etc. etc. It's almost enough to make baseball interesting.

The evolution example involved "debunking" the commonly held idea (by biologists even, I'm not talking religion) that evolution is a continual process toward greater complexity. What else could be more obvious, right? Humans arrived at the end, we're more complex than the rest of the animals.

Well, when you actually look at the numbers, breaking it down evolutionary branch by branch, there is no trend toward complexity (nor toward increased size, or anything else). Evolution appears to be random. Given a species that leads to several other species, half tend to be smaller/less complex, and half tend to be bigger/more complex (bigger and more complex are not going hand-in hand, just showing the two yardsticks by which things were measured). The complexity is just a natural result of the fact that the extreme "right tail" of the bell curve will be more complex.

Gould presents to example of a drunk man staggering in the street. He starts off at the wall, he has a 50% chance of staggering 5 feet toward the gutter, and 50% chance of staggering toward the wall. Given enough time, he'll *always* land in the gutter 30 feet away. Just like you're bound to flip a coin and get 6 heads in a row (about one time in every 32 trials).

He goes on to say that the chances of getting human-like intelligence is super small (my words). He cites several points in the fossil record that show how one or two species going extinct would have meant humans (all mammals) would not have ever been. Also, if it weren't for that asteroid that caused the death of dinosaurs, the tiny mammals would have never had the chance to grow more complex. Dinosaurs commanded the earth for millions of years - the tiny little rodents certainly weren't going to oust the brontosaurus or t-rex.

Anyway, "Full House" was a very good read - shedding new light on ways to look at statistics - that you have to look at the full picture, and not just extremes to come to conclusions.

No comments: