A little disclosure here before I write about Keith Law's new book, Smart Baseball: I spent several years in the shallow end of the pool of baseball statistical analysis. I worked for one team for a short time, did a little writing for the trade magazines, and a few other odd jobs in the business. I'm wasn't an insider but I knew a lot of insiders and man, I wish this book had been available back in the day. It was a lot of fun, but man, reading this book, old hedge wizards like myself would be thoroughly out of our depth in today's game. And honestly? That's pretty cool
Keith Law's Smart Baseball is simply the best baseball book I've read this century. It's clear, rational, funny, and extremely interesting. There are so many wrong turns Law could have taken here; he could have been pedantic or smug or delved so deeply into the technical aspects of baseball's information revolution that it would have rendered the book impenetrable. Instead, it's accessible and informative and a lot of fun to read. If you want to understand the relationship between baseball and baseball statistical analysis, this is the book.
Smart Baseball is broken into three sections. The first concerns traditional baseball statistics and how they present a distorted image of value. It's one thing to say that saves are a terrible stat, but Law presents a compelling case* backed up by just enough data to demonstrate his point. It's bad enough that awards were (and are) given to the wrong players based on reliance on flawed numbers, but teams were making decisions based on bad data, and these decisions were costing teams money and wins.
The middle part of the book is devoted to the current state of the art, the result of the revolution started by Bill James and Pete Palmer and their ilk. The early stat guys, "SABRmatricians," were the ones who questioned the conventional wisdom of baseball and developed mathematical tools to better measure the value of players and strategies. The impact of their work cannot be understated. By the late nineties, more teams than not were making use of advanced stats. And now? Everyone's doing it, and unlike the self-taught enthusiasts of the the turn of the century, today's teams have full analytics departments and proprietary systems for parsing the numbers.
The last section covers the baseball equivalent of the singularity: Major League Baseball's StatCast. The amount of data produced by the in-stadium radar systems, ranging from the relatively simply stuff like "how hard each ball is hit" to near-magical measurements of the spin on a pitch to...who knows? There's more information in there than anyone really knows what to do with yet. Rather than examining existing data with increasingly finer-toothed combs, StatCast opens up a whole new world of data and there's an arms race trying to make sense of it.
It's the "making sense of it" that's the key and makes the whole store so compelling. Anyone can generate statistics; the trick is understanding what they mean and making informed decisions based on that understanding. Law's book is by far the best explanation of the story of how analysis has changed the game for the better that I've ever encountered.
* Not that this is a terribly difficult case to make when you're talking about saves...