Moneyball by Lewis, Michael (mobile ebook reader txt) 📕

the element of luck. Everyone in baseball knows how much luck is involved in the game but they all say, ‘The luck evens out.’ What AVM was saying is that it doesn’t. It’s not good enough to say, ‘Aw, it just evens out.’”

An insight born in the financial markets took root in the minds of a young man who would soon have the power to put it to use inside Major League Baseball. Not long after Billy Beane had hired Paul DePodesta, in 1998, Paul persuaded Billy to hire AVM Systems. “They were still interesting to me,” Paul said, “because they weren’t churning conventional statistics in unconventional ways, which is what everyone else does.” AVM Systems was a luxury only a rich team could afford but that only a poor team, desperate for any edge, would think to use. Billy and Paul used the AVM system for a couple of years and then, to save money, copied what AVM did. Once Paul finished replicating the parallel world of derivatives, he and Billy could begin to answer more accurately the question about Johnny Damon’s defense.

Every event on a baseball field Paul understood as having an “expected run value.” You don’t need to be able to calculate expected run values to understand them. Everything that happens on a baseball field alters, often very subtly, a team’s chances of scoring runs. Every event on a baseball field changes, often imperceptibly, the state of the game. For example, the value of having no runners on base with nobody on base and no count on the batter is roughly .55 runs, because that is what a baseball team, on average, will score in that situation. If the batter smacks a double, he changes the “state” of the game: it’s now nobody out with a runner on second base. The expected run value of that new “state” is 1.1 runs. It follows that the contribution of a leadoff double to a team’s expected runs is .55 runs (1.1 minus .55). If the batter, instead of hitting a double, strikes out, he lowers the team’s expected run value to roughly .30 runs. The cost of making that out was therefore .25 runs-the difference between the value of the original state of the game and the state the batter left it in.

But those calculations really only scratch the surface of the problem. If you want to strip out the luck and get to a deeper understanding of the value of a player’s performance you have to pose the baseball equivalent of existential questions. For instance: what is a double? It really isn’t enough to say that a double is when a runner hits a ball and gets to second base without a fielder’s error. Anyone who has seen a baseball game knows that all doubles are not alike. There are doubles that should have been caught—just as there are balls that are hit that should have been doubles but were plucked from the air by preternaturally gifted fielders. There are lucky doubles and unlucky outs. To strip out the luck what you need, really, is something like a Platonic idea of a double.

A set of Platonic ideas is one of the gifts the Wall Street traders gave to Paul DePodesta. The precision of the AVM system, copied by Paul, enabled him to think about every event that occurred on a baseball field in a new and more satisfying way. Any ball hit anyplace on a baseball field had been hit just that way thousands of times before: the average of all those hits was the Platonic idea. Call it a line drive hit at x trajectory and y speed to point #968. From the ten years worth of data, you can see that there have been 8,642 practically identical hits. You can see that 92 percent of the time the hit went for a double, 4 percent for a single, and 4 percent it was caught. Suppose the average value of that event is .50 of a run. No matter what actually happened, the system credits the hitter with having generated .50 of a run, and the pitcher with having given up .50 of a run. If Johnny Damon happens to get one of his trademark jumps and makes a sprawling catch, he is credited with saving his team .50 of a run.

The beauty of the value of that hit (or catch) was that the game gave it to you; the game told you how valuable every event was, by telling you how valuable it had been, on average, over the past ten years. By listening to what the game told him about the value of events, Paul could take every ball hit between in the area broadly defined as center field and determine its “expected run value.”

Which brings us back to Johnny Damon. Over the 2001 season many hundreds of balls had been hit by opponents of the Oakland A’s in the vicinity typically covered by the center fielder. By totaling up the outcomes when Johnny Damon was in the field, and comparing them to the average, Paul was able to see how many runs Damon had saved the team. He was also able to estimate how many runs Damon’s likely replacement, Terrence Long, would cost the team. Some of this you could see with the naked eye, of course. You could see Johnny Damon break the instant the ball left the bat. You could see Terrence Long freeze, or even take off in the wrong direction, when the ball was in midflight. You didn’t really need Wall Street traders to tell you which one was the better center fielder. The system born on Wall Street simply helped Paul to put a price on the difference. There was no longer any need to guess. There was no need for gut instinct, or conventional fielding statistics. The total cost of having Terrence Long, rather than Johnny Damon, in center field was fifteen runs, or about a run every