“Momentum” in Baseball

28 May 2005, 1828 EDT

As a baseball fan, I watch a fair number of games during the regular season, most of them on television (ballpark prices, even for the cheap seats, prohibit more than about one family outing a month to see a game live). A lot of the commentary by the sportscasters is silly, although the anecdotes told by former players and the locker room gossip that reporters have managed to pick up are sometimes interesting. But the analysis of what is going on in the game or in the season as a whole is, by and large, silly, since it deploys questionable analytical strategies and dubious evidence, such as the reverence that many commentators have for strategies like the sacrifice bunt — strategies that are only rarely reasonable moves to make during a game. So what we often get from sportscasters is nothing more than speculation, sometimes speculation informed by experience, but more often speculation informed by some unsupported general sense of how the game should be played.

The worst abuse of technical language that I have encountered during these sportscasts, though, involves the word “momentum.” Time and time again I hear assorted commentators discussing the “momentum” that a team has, as though “momentum” were an attribute of a baseball team and could be reliably utilized as a basis upon which to make predictions about the future performance of that team. “Momentum,” in these folk analyses, is treated as a mechanism that explains an object’s trajectory through space, with the object being the team and the space being the regular baseball season. And partisan commentators regularly admonish their team to “keep the momentum going” or to “get some momentum up” by winning several games in a row, a locution that presumes that a team could somehow alter its momentum through more or less deliberate action and thus produce more wins.

This is a profoundly misleading notion, and the precise ways in which it is misleading might serve as an object lesson for other attempts at predicting the outcomes of social processes.

Momentum, strictly defined, is a measurement of an object’s motion that defines, in effect, how hard it is to stop that object. Momentum is calculated by multiplying an object’s mass by its velocity, where velocity is a scalar quantity that describes both the speed and the direction of motion.1 Two things are highly relevant here: first, momentum is a description of motion, and only “explains” it in the technical sense of providing a way to calculate an object’s position at time t+1 given its position at time t and a knowledge of its momentum at time t; and second, like any analytical operation, a prediction based on such knowledge has an enormous ceteris paribus clause: the model deliberately excludes a lot of context, and presumes that it isn’t relevant to the outcome in question.

What do these qualifications mean in practice? The first qualification makes something of a mockery of the idea that a team can improve its record by changing its “momentum,” since momentum isn’t so much a causal mechanism producing outcomes as it is an element (and only one element) of an analytical system explaining outcomes. In order to produce outcomes, we’d need to know what mechanism or mechanisms accounts for whatever “momentum” a team has at a given point in time, so that we could modify that mechanism directly. “Momentum” is, at best, an intervening variable.

And that’s only at best, because the second qualification — the ceteris paribus clause associated with all attempts to predict based on a model — renders the notion of momentum pretty much unusable for social phenomena. Think for a moment about what is presumed when someone says that a team has “momentum”: that baseball teams are akin to solid physical objects (like the baseballs that they toss around the field), that their direction of flight through the air is as definable as that of any physical object, and that the parameters of the system within which the objects are flying are more or less parametrically fixed over the whole course of the period in question.

Take these each in turn:

Are baseball teams like solid physical objects? Well, given the frequency of trades, roster moves, line-up shuffles, and the like, I’d have to say “no.” Calculating the mass of a baseball team — an essential part of the momentum formula — is thus pretty much impossible, and would be roughly akin to trying to deal with the flight of a baseball whose mass shifted while in flight. And even if the overall mass (however calculated) remained constant, the continual adjustments to the roster and the lineup would make it impossible to precisely describe the velocity of the team, since the simplifying assumption (that mass is concentrated in the center of an object) ordinarily used to approximate such things would no longer be empirically valid.2

Is direction of flight precisely calculable? Well, with a baseball moving through the air, in principle, yes, because we can set up radar guns and laser trackers and the like. And once we have a general equation describing the path of a ball in flight, we can use that as a basis on which to predict what is going to happen the next time a ball is thrown, assuming for the moment that nothing fundamental will have shifted by then (such as a hurricane suddenly coming in, or gravity in the stadium altering, etc.). But with a baseball team, we have an epistemological endogenity problem, in that players and managers (and owners … especially intrusive owners like George Steinbrenner, czar of the Yankees) read the analyses offered and can factor those analyses into their ways of proceeding. This means, in effect, that the direction of flight of a social object like a baseball team — or the global political economy — can’t actually be precisely measured. We can describe the instantaneous rate of change of the object’s position3 at some past point in time, but describing it in the present always runs the risk that this very description will alter the object’s velocity, as the team takes concrete steps to change the direction of its flight.

Antonio Gramsci, of whom I am not the biggest fan, nonetheless put his finger on it when he pointed out that predictions in the social world were impossible because of the ways in which such predictions interacted with the people whose trajectory was being predicted. Likewise, Isaac Asimov nailed this problem in his Foundation series of novels when he made it a condition for the success of succesful predictions of the human future that the subjects of the calculations be kept ignorant of the details of the calculations. Whatever we are doing when we make predictions about the social future, it isn’t the same thing as physical scientists do when extrapolating the trajectory of a physical object.

Do the parameters of the system remain intact over time? Well, in baseball, I think the answer has to be a qualified “yes” — yes because there are whole organizations set up to prevent the parameters from shifting, and qualified because some of the fundamentals (like the height of the pitcher’s mound and the precise boundaries of the strike zone) have in fact changed over time. So if one could overcome the other problems, which I don’t think that you can, there might be a prima facie case for extrapolating from a team’s past performance to its future performance, and giving a notion like “momentum” some (pardon the pun) real analytical weight.4

Hence: predict baseball outcomes with extreme caution, and avoid notions like “momentum.” Specifying actual mechanisms that produce wins is a much more analytically defensible way to go.

That having been said, if the Yankees keep playing the way that they have been the past few weeks, they’ll win the AL East again, and I doubt that the wildcard will come from the East this year; the AL Central has so many bottom-feeding teams that I suspect that both Minnesota and Chicago will amass records sufficient to get them to October.

1Even though most baseball commentators only use “momentum” to describe a team that has strung together a series of victories, there is no logical reason why one couldn’t similarly say that a team on a losing streak had “momentum” in the opposite direction. That is, if you wanted to use “momentum” to describe such things in the first place.
2Imagine a baseball the interior of which was actually composed of a series of small metal weights, the position of which was affected by a magnetic field generated by a device at the center of the baseball. Imagine further that this magnetic field could shift. Even though the overall mass of the baseball would remain constant, its precise character could be shifted in small ways by redistributing the weights inside of the ball; moving the baseball’s center of gravity would affect the character of its velocity, and hence the character of its momentum. Now, apply this to a baseball team composed of many more moving parts.
3Technically, this is what velocity is: the first derivative of position.
4The closest thing I know of that does this is the so-called “pythagorean average” formula invented by sabremetrics guru Bill James. But this formula differs from a “momentum” formula in that it incorporates actual mechanisms that generate wins and uses those as a basis to forecast a team’s performance.