This one is dedicated to the memory of my father-in-law, himself a biochemist. I once tried explaining baseball and sabermetrics to him (he was from Russia). He thought it was nice that I had such an interesting hobby. He will be missed.

Clubhouse chemistry. The fact that everyone in the room believes in everyone else. Except Smith. The fact that this team is composed of a great bunch of guys who are there for each other and support each other. The fact that no one could ever ask for a better group of teammates. The fact that we just won the World Series.

The chemistry question is one that flummoxes data-driven investigation, mostly because there just isn't any direct evidence to test the hypothesis. Can an ill-defined and nebulous concept like "chemistry" among the people on the team have an effect on their performances on the field? Performance can be measured, but how on earth does one actually measure chemistry? It's not like MLBAM is putting out BESTFRIENDf/x any time soon. (MLBAM, if you're reading this… I'm just saying.) Outsiders don't have access to the clubhouse, either directly or indirectly. Players and media usually have an unspoken agreement that what happens in there stays there. About the only time that outsiders even see a locker room is when players are spraying each other with champagne after clinching a tie for the second Wild Card, entitling them to a play-in game to get into a play-in game. No one interviews the team that lost 100 games. Maybe everyone on that team liked each other too. And anyway, even if they did interview them, there's a standard answer that everyone knows how to mouth.

There are plenty of reasons to be skeptical of the claim that chemistry influences performance. One commonly cited idea is that the arrow is pointing in the wrong direction. Chemistry doesn't influence performance. Winning puts everyone in a better, friendlier mood. (More on that a little later.) Another is the selection bias of the "proof" that's often offered. Players get the chance to proclaim how joyful things are only when they've won something, which gives the illusion that winners live in happy locker rooms. And even if people can barely stand each other, there's the old rule that if you can't say something nice, don't say anything at all. These are all legitimate points that the available data are very biased. We can't scientifically accept the claim that chemistry influences performance.

But we can't dismiss it either. Often, I hear well-meaning sabermetricians deny the claim out of hand. How can anyone possibly buy into this stuff? There's no proof! But if we're going to be fair, the chemistry-performance link falls into the "reasonable hypothesis" category. We just don't know either way. It might not be true, but it at least passes the "if you say it out loud, you don't sound foolish" test.

In the absence of direct evidence, we're forced to look at (very) circumstantial evidence on the matter. In favor of the hypothesis, there's the fact that an entire field of study exists (industrial/organizational psychology) that looks at the evidence on "workplace environment" and productivity in other types of jobs. The general evidence is that the structure of the workplace and the interpersonal relationships between workers make a difference. A baseball team is just another workplace, right? On the other side, a few weeks ago, when I tried to find an effect for one specific type of workplace chemical agents (that just sounds wrong…) that teams tend to employ, the "veteran leader", there was no major effect that blanketed the whole team, although there might be some effects on individual players that we wouldn't be able to detect.

It's tempting to think that it's impossible to measure team chemistry. I personally think "impossible" is the wrong word. It's not a physics problem; it's a logistics problem. In fact, if I had the access (and the budget) to do it, measuring clubhouse chemistry wouldn't be all that hard. To get a big enough sample size that would produce meaningful results, I'd probably need access to a majority of team clubhouses over multiple years, not to mention the cooperation of the players in answering my questions. This isn't likely to happen (again, if anyone's reading… just saying…) but I think it's interesting to look at how it could be done, since it can tell us some interesting things about what actually comprises "team chemistry" and what effects it might realistically have. In what follows, I present a research project that will never be fully realized. But you might just learn something from the exercise.

Warning! If There Were Gory Mathematical Details, They Would Go Here!
First of all, let's talk about what we mean by chemistry. There are 25 players on a team, and each of them experiences the clubhouse in a different way. There is no team where all players are BFFs with the other 24 players. It wouldn't be a horrible thing, but if I found a bunch of guys like that, I would politely ask the cult members to direct me to the baseball stadium that I apparently mistook their compound for. What we're looking for is a well-functioning and well-connected social network.

There are a couple of different ways to measure the development of a network, again assuming massive (and unrealistic) amounts of access and cooperation. The easiest one is what's known as a peer nomination strategy. You give each of the players a sheet of paper and ask them to write down the five or so players on the team with whom they feel they are most friendly (or some variant of that question). If you want to get fancy, you can ask them to rate that closeness on a scale of 1 to 10 (or 20 to 80). You can then map out who is friendly with whom and figure out how many links on the chain it would take to get from Smith to Jones. If Smith hangs out with Lewis, and Lewis is also friendly with Jones, they are two links apart. If there is conflict between Smith and Jones, they are more likely to solve it if they share friends in common, both because Lewis might act as a peace broker, or more to the point, Lewis will be annoyed at being in the middle of a fight and whip the two players into shape.

In an ideal clubhouse, there would be at least some sort of link between everyone on the roster. The problem comes when there are players (or groups of players) who are completely isolated from the rest of the team, something that would show up on that social map. It's easy for humans to fall into cliques. On a team, lines might be drawn along pitchers vs. position players, English vs. Spanish speaking, religious vs. non-religious, or about 1500 other possible granfalloons (a billion points for that reference). Those aren't impossible barriers to break down, but if no one makes the effort, then it leaves open the possibility of conflict and no good interpersonal path for solving it. In other words, bad clubhouse chemistry. Everyone doesn't have to be friends with everyone else, but well-formed networks do offer a counterweight to festering conflicts.

The cool thing for the quantitatively-minded researcher is that there are numerical methods for looking at these social maps. You could look at the average distance between players or look at how many factions are in the clubhouse. You can look at the general ratings, subjective as they are, and see if this is a clubhouse where there are a lot of 50 relationships or a lot of 80 relationships. A researcher could start by looking at how individual players fit on the map and check it against their performance, and then do the same at the team level. If a well-connected clubhouse helps players, we would be able to prove it. If it's true, then that guy who is a good people person, makes friends with everyone, and has a second career as a diplomat awaiting him but also hits .230 suddenly has more tangible value beyond his on-field WARP. It's possible that being that link between guys who are fighting can be quantified into runs. The critique of the WARP model (and sabermetrics in general) that I've heard most often is that it doesn't consider the value that a guy adds in the clubhouse. Well, here's a good start on how we might address that. Want a sabermetric Nobel prize?

On the question of whether happy players hit the ball further, a researcher can ask simple questions like "How happy are you on a scale of 20 to 80?" before the game, after the game, and the next day when the players get to the park. It would take a little bit of time to get what each player's baseline level of happiness is (because this is a subjective scale, some guys will say 80 when they are only moderately happy and some will save 80 for being deliriously happy), but that's just a matter of getting a good baseline. Once that's been established, though, a researcher could see whether a player being above or below his baseline happiness rating helps or hinders his performance. And the same for the team. Maybe hanging out with a bunch of chipper chaps makes even the most dour and awful starter pitch like Justin Verlander.

The method of looking at happiness before and after the game would shed light on the question of whether happiness leads to performance or performance to happiness. However, I'd personally expect the arrow to point in both directions. This is neither logical nor statistical heresy. There are plenty of cases where two variables feed off one another. You feel better so you have some energy to get up and work out, and that makes you feel even better. These are called reciprocal effects and can be modeled with either some good longitudinal regressions or with a structural equation model. Again, with the right data, it wouldn't be that hard.

And so if baseball would grant that level of access, we'd at least have quantifiable measures of social connectedness in the clubhouse and player mood. Combined with some outcome measures on the field, we can pull together a few hypotheses to test whether those post-World Series speeches are actually true and whether it made a difference. We might also look at which teams (and players) are emotionally the most hardy in the face of an 0-for-5 night, and perhaps which teams (and managers?) are the best at not getting too high after a win or too low after a loss… and whether that one matters either.

While I'm in the clubhouse, I might look at a couple other markers that would promote good chemistry. How often do players converse with one another after games, and for how long? (Yeah, I know that would be creepy to actually measure.) How long do they hang out in the locker room after a game? These can be indicators of a clubhouse that no one likes to be in. I might also look for some structures that would promote chemistry. Does the team have a structure (a kangaroo court?) for settling grievances that arise? Does the team have an identity that everyone buys into? Say what you will about the Rays’ themed road trips being silly, but they get everyone on the team to buy into something that they all do together. A manager must be one part field tactician and one part summer camp counselor.

And when you add it all up, it sounds like a good measure of team chemistry. It isn't really that hard to measure, if you can get the right access, and then it's just some numerical gymnastics to figure out whether there's a link to performance. What's important to know about the things that I've laid out is that they all have been shown to promote a good workplace environment, and that generally increases productivity. Now, in a baseball context, the effects might not be huge. Only in Hollywood can a ragtag bunch of rejects turn into world beaters just by trusting in each other. But I think that given what we know, the most logical guess would be that there would be some effect of chemistry, just maybe not as exaggerated as it seems in the warm glow of a late October night.

But if you ever wondered why even the "smart" teams pay so much attention to such a squishy thing as "team chemistry" consider this: the effect might be small or maybe even non-existent, but the cost of trying to optimize it is miniscule. It might be the cost of some cheap props from a dollar store. Even if a loose clubhouse adds half a win, if the cost is 25 bucks, that's a deal. And even if the interventions fail, they probably won't make things worse. So why not try?

And maybe the effect is a little bit bigger. Right now, we don't know, but I think we as sabermetricians do ourselves a disservice if we assume that chemistry doesn't matter. And maybe some day, we'll get a chance to prove it one way or the other.