Sunday, February 6, 2011


Those who were upset with the "swipes" taken here and there in BBBA at the various folk who've taken the concept of sabermetric research and turned it into a commodity will probably see the headline here and run for the hills.

Call me Ishmael, not "Rajah"...
All the better. One is freed by one's ability to write without the cavils of groupthink...

Back in the late 1990s, we predicted what would happen to the field of sabermetrics. We suggested that it would, in many ways, become indistinguishable from the mainstream reportage that most of the folks at Usenet disparaged. In short, that "the movement" (as some folks liked to call it) would be co-opted.

Let's take a look at what's happened since.

--The Baseball Prospectus folks got a gig with Sports Illustrated, and, on the heels of this mainstream success, transitioned themselves into an actual business. Over the past decade, they have routinely imported and expelled writers from their organization in a fashion that resembles broadcast media (the lines of which, as we all know, have become increasingly blurred in the past fifteen years).

--Many of those folks who've been through the Prospectus' revolving doors have used it as a springboard for other efforts. While these efforts purport to "correct" the viewpoints and theoretical underpinnings of the evolving "research," they most often rely on similar feats of reverse engineering that still plagues the work of Baseball Prospectus (and others).

--Michael Lewis' book, Moneyball, which barely engages actual sabermetric theories, became a best-seller in large part due to its author's talents at portraying the complicated but charismatic personality of Oakland GM Billy Beane, whose success from 1999-2004 was a rallying cry for the "neo-sabermetric" movement that culminated in a large percentage of baseball teams hopping on the "numbers" bandwagon.

Yes, it's really this simple!
--This, in turn, has produced a feeding frenzy of activity that has had mostly negative effects. While it produced a few new jobs, it also created a different climate for the entire endeavor and has injected an even higher amount of venom into the process than was ambient in the late 1990s.

The most unfortunate outcome, of course, is the fact that research methods and products have become commodities first and instruments of knowledge/understanding second. The "proprietary" model for "advanced metrics" has created an odd situation wherein several basic measures that use the same name (Wins Above Replacement) are now at odds with one another.

And just how would the "paradigm"
of "advanced metrics" have "shifted"
if it'd turned out that this was what
Voros McCracken looked like??
SUCH an irony is not lost upon yours truly, who predicted this outcome at the time. Recent discussion threads at the Baseball Think Factory featured a profile of Robert (Voros) McCracken, the analyst responsible for the basic insight that has driven an entire suite of increasingly murky equation-based modeling that attempts to recalibrate a pitcher's performance to a set of parameters that simulate what that value would be if all the biases involved with defense are adjusted.

These super-models involve a series of proxies for data that either doesn't exist in the form needed in order to be measured directly or that has been regressed into relationships that are at best educated guesses for how each component might affect the other.

The complications of these are enough to clearly warrant a kind of "think tank" approach (as I noted in one of those threads, and in an earlier post here), but what we're finding to be the case is that the most forceful personalities (to be distinguished from the best minds, as an historical examination of most organized theoretical inquiry in the past hundred years will divulge) often dominate the process and become overly prominent.

I suggested in one of those threads that major league baseball should recognize the fact that "advanced analysis" is something that should be taken out of the hands of individual teams. What's happening with the competitive, capitalist model is that it's producing a similar type of meritocratie maudit that I was taking Rob Neyer to task for in an earlier post: there is a kind of systematic distortion effect that comes into play when a meritocracy becomes rigidly pyramidal. The received values, precepts and rituals of that group forms an elite, which then has a strong tendency to fold in on its already-received wisdom.

Since baseball is first/foremost a business, it's easy for the folks that run it to work a kind of end-game on this new class of "expert workers." Under such a scenario we're likely to see a conglomeration of thirty small fiefdoms, each empowered with meagre resources, all trying to catch lightning in a bottle.

That phase is likely to last for ten, maybe fifteen years. The fiefdoms will be increasingly threatened by (and forced to creatively adapt to, as in any bureaucratic "evolution") outside technology companies, who will further muddy the waters by inventing measurement devices that, in absence of anything resembling systematic thinking, will create tools that produce data which will mostly be disconnected to the current paradigm of uber-modeling. (In fact, this is already happening, and we can see the cleavage--no pun intended!--between the research approaches as these tools produce "data.") This disconnection will create additional conflict and chaos between the groups vying for dominance in the "baseball theory business."

The problem--no one is doing a sanity check on any of this. The musical chairs of neo-sabermetic careerism have accelerated to the point where the accompaniment is something more suitable for a mosh pit than a ballroom. Without a true meritocracy, one in which the best and brightest get measured in a systematic, unbiased way, the field will spiral further out of control, with more competing versions of the same measures and no reconciliation of the grey areas that come into existence as a result.

The most damaging aspect of the "neo-sabermetic" movement is its promotion of a particular style or brand of baseball as a result of its value modeling. Its "neo-meritocracy" of fiefdoms isn't likely to even recognize this as a problem. The irony of complex mathematics actually contributing to the institutionalization of a "dumbed-down" version of the game is lost upon those who are using these precepts to find a way "inside."

To demonstrate the point, let's look quickly at two charts created from the "breadth of peak" discussion in an earlier post. The first chart looks at what the league-relative isolated power (ISO) and walking percentage (BBP) for players with six-year peaks of 130+ (as measured by OPS) look like over time. What it tells us is that these hitters are becoming more uniform in how they excel. Their relative ISO has been contracting, while their relative BBP has been expanding.

Why is this happening? The second chart (below) gives us the answer. When we look at the ratio between ISO and BBP relative to the leagues from the deadball era to the present, we see a series of oscillations in ISO over BBP driven by peak performers. There are two especially notable peak points (as shown in the blue line)--1930 (which won't be surprising) and 1961 (which, in keeping with the conjectures of Brock J. Hanke, was a touchstone for the Lords to back off from the emphasis on home runs that had made big strides in the 50s). There is another peak in the late 90s, but it is not nearly so dramatic as the others. Why is this?

The red line--the MLB average ISO/BBP ratio from 1906-2010--provides the explanation. The distance between peak and league average stays pretty constant in the first sixty years; there's a brief convergence in the mid-60s as teams expand power past the 3-4-5 slots in the lineup to cover lower overall batting averages, but what's striking is that the next twenty years start to produce a convergence in the characteristics of peak hitters and the MLB average as a whole. The power surge in the late 80s comes from a resurgence from non-peak hitters, and this shifts into overdrive in the 90s, moving upward at a rate virtually unprecedented in all of baseball history, pushing itself to an all-time high in 2004.

In short, the style of hitting being practiced by top players and average players has converged. This was the "hidden" impact of sabermetrics in the nineties, and those who are now employed by MLB teams are subconsciously working with this ambient condition already in place.

Now, one could argue that this is a good thing. But such a discussion--with its "pros" and "cons"--cannot really happen when the best and brightest are not working together, but as "competitors" who are owned by their respective franchises. Without a centralized function for these types of questions, the "false meritocracy" will be unlikely to address issues that are quite possibly critical in their impact on the future of the game.

We are now reaching a time in the history of "advanced baseball analysis" when a centralized "think tank" is becoming necessary. How to bring that into being, however, is a tricky prospect. Some of the folks who have been entrepreneurs, of course, are going to have to become part of that organization. But it's clear that, whatever approach is arrived at for populating this group, it will need to have control over the development of technological tools used for the purpose of creating new data and new measures.

This is an issue for the next commissioner of baseball to address; hopefully that will be at or near the top of the agenda in the next 3-5 years. The looming chaos of the "neo-meritocracy," inherited from the unfortunate sixty-plus year trends in government technology, needs to be reversed before it becomes a fait accompli.