The triumph of Moneyballby Pizza Cutter
November 10, 2009
Is there a word in the baseball lexicon that can start an argument faster than Moneyball? It’s not really a word, if you think about it, but then neither is the only other thing that springs to mind: DH. It’s odd that a book not called “The Bible” could promote such arguments, but an actual counter-revolution formed (or perhaps simply came to the forefront) at the thought that anyone would go near the game of baseball with a laptop and a spreadsheet. Thankfully, Moneyball was just a silly idea that wasn’t even welcome in San Francisco(!) It had to go to Oakland to get some traction. Or so people thought.
One of the oft-parroted (and oft-misunderstood) lessons of Moneyball was that on-base percentage (OBP) was the statistic by which to rate a player. In fact, Moneyball made the case that the A’s success was built on an understanding that OBP had two properties. It was a more effective way than batting average to rate players, and OBP was inefficiently priced in the free agency market. Detractors charged that chicks dig the long ball, but nerds dig the walk. But something interesting happened. A few more teams publicly embraced Sabermetrics within their front offices, with a few hiring well-known Sabermetricians to be in-house stat-heads. While this wasn’t a majority of teams, it was a notable minority. Still, traditionalists scoffed and wondered why none of these teams (meaning Oakland) had yet won a World Series.
Did Moneyball really have an effect outside of Oakland and the handful of teams that embraced Sabermetrics in the following years? Were the lessons of Moneyball taken to heart league-wide? The surprising answer is “Yes, and in a much more powerful way than you might expect.” The way to tell whether a man believes something is if he’s willing to “put his money where his mouth is.” In baseball, that’s rather literal. Teams buy the service of players in a (mostly) open market. How much money they are willing to commit to a player tells a lot about what they think of him. And what drives those salaries tells a lot about what the market as a whole thinks about what makes a player valuable.
I took 11 years worth of data, 1997-2007, which corresponds to five years before the release of Moneyball in 2002 to five years after. What I wanted to find was what statistics appeared to be driving the salary market during those years. I selected all hitters who had more than six years of MLB service (as dated from their debut year) during the season in question. This weeds out the players who under the new labor contract were in the “slave labor” years of their careers, prior to when they could file for free agency. (The old agreement didn’t have the same structure, but the nice thing about six years in the majors is that it makes everyone in the study a well-known quantity.)
At first, I looked for correlations within the year between salary and performance. But then I remembered a conversation that I had with my brother on the way back from a game. He has a master’s degree in finance, but he pointed out to me that owning a baseball team is a really awful business model. Almost all of your costs are fixed (and almost all of your income is variable). Teams sign players before they get the performance out of them, and players get the same amount whether they play like MVPs or they get hurt on Opening Day. GMs are looking at the past few years of performance and hoping that history repeats itself. So, I began looking for correlations between salary and performance from the few years earlier, which is what information the GM had on hand when the salary was assigned. Correlations were generally strongest two years prior, although usually only slightly above one and three years prior. So, I looked for correlations between salary and whatever stat was of interest two years earlier. To qualify, a batter had to have at least 250 plate appearances in (year–2).
First let’s look at a graph of how well a player’s OBP from two years earlier tracked his salary.
The year 1997 may be an outlier in this case, but the correlation between the two was .31 in that year. We see that in the late '90s and early oh-ohs, the correlation danced between .40 and .50. In 2001, one year prior to Moneyball, it was at .44. By 2004, it was .64. The strength of the correlation (as measured by R-squared) about doubled. Coincidence? Maybe. But maybe, just maybe, the people who actually make the decisions in baseball actually read and accepted the conclusions in Moneyball. (A small aside: Batting average was always below OBP in its strength of correlations. Ideas that front offices were pricing batting average are not actually justified. If anything, AVG did a horrible job tracking the market.)
The tail end of that graph is concerning, as we see the correlation beginning to fall off. Perhaps Moneyball was a fad. It had its couple of years in the sun, and then… well, Jean-Luc all good things must come to an end. (Sorry.) Let’s take a look at what happens when you look at two other stats that really drove the market, home runs and (sorry for the four letter word) RBI. Those are just the raw numbers as in “Smith hit 35 HR last year and drove in 110.”
Note that in the years before Moneyball, HR and RBI clearly drive the market much more clearly than does OBP. By 2004, the jump in OBP’s popularity had pulled it even, partly because HR and RBI fell in their correlative power. In 2005, OBP was actually the better correlate of salary. Chicks may dig the long ball, but apparently nerds were running the front office of your favorite MLB team. Look what happens after 2005 though. There’s a general downward trend for all three stats. It’s likely that OBP did have its day in the sun, but why would HR and RBI, so long dominant, also fall?
Over the past few years, we’ve seen the proliferation of a number of advanced statistical techniques, whether total value functions incorporating offensive performance and defensive prowess, or context adjustments such as replacement level or park adjustments. My guess is that if I could pull together a database on the subject, these advanced stats would show an upward tick in their correlative power with salary. OBP was the beginning. Now teams are into the real stuff.
So what does it all mean? It means that whether by cosmic accident or intelligent design, the principles espoused in Moneyball seem to have won the day in the only place where it really matters: the front offices of Major League Baseball.
Pizza Cutter has a Ph.D. in clinical psychology, and uses all that education not to rid the world of depression and anxiety, but to study baseball. His work has appeared at Statistically Speaking.