My wife and I have been married for seven and a half years. We dated for four years before that. There are days when it's eerie how in sync we are. We've gotten to the point where someone will say something and we’ll both look up and smile knowingly at each other because we’re both aware that the other's mind just went to the same obscure song lyric from 15 years ago. Yeah, I think we have some chemistry going.

And then there's the chemistry I have with the man who probably controls my life even more: the best boss in the world, BP's editor-in-chief, Ben Lindbergh. He sent me an e-mail last weekend which contained a couple of quotes and an idea for an article on the subject of team chemistry. Since I wrote about chemistry a couple of weeks ago, he thought of me when he saw them. Remind me to send him some flowers.

It's that odd time of year when players are reporting to camp after having been away for a few months. I'm sure it's a nice time to reunite with old buddies whom you haven't seen all winter but whom you spent six months with in rather close quarters last season. There's surely a lot of high-fiving, back-slapping happiness as guys get to catch up with one another. And it must be kinda tough for the new guys. After all, they walk into a clubhouse where most of the people know each other, and they have to be the new kids in school.

We've also reached the awkward point in camp where teams are starting to play games but most players aren’t at maximum effort, so there's not much to talk about other than the obvious five questions about each team that often have five obvious answers. It's noteworthy when anyone says anything remotely interesting. Brandon Belt was asked about the Giants' NL West rival, the Los Angeles Dodgers of Los Angeles, and how they have welcomed quite a few new faces into the fold recently. Belt stated, "You can't buy chemistry." Evidently he figures that since the Dodgers have a lot of new faces on the team, their on-paper talent may not have had time to jell yet, and so they might not be as fearsome as their projections indicate.

Meanwhile, in Nationals camp, there was this dispatch from Washington Post writer Thomas Boswell:

In an itinerant mercenary sports world, the Nats are recreating something from a previous age — as much as the era permits. They want to be a club that stays intact year to year in its core personnel. That way, they accumulate knowledge as a group from season to season. More important, they want to embed a clubhouse culture in which everyone aspires to ferocious focus.

I'll pass on figuring out what "ferocious focus" means, but all this talk brings up an interesting issue. Is keeping the same group of players together somehow good for a team's performance? Does the third baseman being able to complete the left fielder's sentences or the catcher knowing what was the shortstop's favorite band in 10th grade somehow improve a ballclub’s record?


Warning! Gory Mathematical Details Ahead!
First, we need a measure of how much turnover a team has experienced. I looked at all players (both pitchers and batters) from 2004-2012 who played for a team in at least 20 games. To do that, you have to be around for at least a month, and a month cooped up in close quarters with anyone will drive someone insane. I then looked to see whether the player had played for the team in the year prior. (Note: this does have a few unfortunate side effects. Starters who get hurt and miss a dozen starts might end up with 19 games, while a September call-up who gets a three-week audition might qualify. Such is life.) A team's carryover ratio was the percentage of players who were around for 20 games who also were with the team the year before (even partially). I also coded for whether the player was a holdover himself or a new kid on the block.

Then I selected all hitters who had 250 PA in both the previous season and the current season. I looked at various rate stats, including strikeouts, walks, HBP, singles, extra-base hits (2B + 3B), home runs, and outs in play per PA. I also looked at AVG/OBP/SLG.

I tried two methods. I first used a method in which I predicted the current year's rate, controlling for last year's rate and then including the carryover ratio in the regression. I also included the dummy variable of whether the hitter was a holdover or not, and the multiplied interaction between the two. (For the initiated, that's a dichotomous moderator model.) Surprisingly, there were significant findings for HR rate and outs in play.

For HR rate, I looked at four possible scenarios: The team had a carryover rate of 50 percent (high turnover) or 90 percent (low turnover) crossed by whether the player was a holdover from the previous year or whether he was new in these parts. I assumed that he was roughly a league-average hitter, so I also assumed that he hit a HR in 2.5 percent of his PA last year. These were his projected (very age and park unadjusted) HR rates for the following year.


High Turnover

Low Turnover

Player was here last year



Player was not here last year



For outs in play, I assumed a base rate of 50 percent.


High Turnover

Low Turnover

Player was here last year



Player was not here last year



An interesting set of findings. The worst outcomes happen for a player who is new to town and who is walking into a room full of guys who played together last year. It's hard to be the new kid in school. However, when there are a lot of new people around, the new guys actually do better.

The effect sizes here may not appear huge, but for those who are holdover players, even going from high turnover (2.46% projected HR rate) to low turnover (2.59%; a difference of 0.13%) over 700 PA is just shy of one extra home run. But since that effect can theoretically apply to all the players on the team, you might estimate that the team could pick up a handful of extra home runs that otherwise might have been outs. The advantage gained from this sort of switch is something on the order of a run and a half. If you can gather seven of those together, it's worth a win.

I did something similar for pitchers using the same basic format, looking at strikeout, walk, HBP, and HR rates for pitchers with 250 batters faced in each affected year. That one came up empty. I couldn't find any significant effect for how well pitchers performed.

Moving to the team level, I looked to see whether there was something to the thought that team cohesion might make a team better than it looks on paper (or in Brandon Belt's formulation, that the Dodgers will have a price to pay for so many new faces being around.) Thanks to some intrepid work by BP intern Andrew Koo, I took PECOTA's preseason projections from 2004-2012 and compared them to what they actually achieved. The carryover rate and the amount that teams varied from their preseason PECOTA projections had a correlation of .03. In other words, no relationship. Sorry, Brandon.

But perhaps being a cohesive unit helps a team win the close games? As Hollywood has told us over and over, the key to winning is commitment to each other, courage, and character. Well, the correlation between carryover rate and record in close games made it only to .08. Sorry, Hollywood.

But is it Chemistry?
For those of us re-joining us because they averted their eyes from the gory details, I found that there is evidence that the amount of turnover that a team experiences has an effect on a hitter's performance, to the tune that—under ideal circumstances and assuming that we have a causal relationship rather than just a correlation (more on that in a minute)—a team might leverage a few extra home runs from roster consistency. However, there was no evidence that pitchers were affected. Nor was there any evidence that teams performed better or worse than they looked "on paper" due to high or low turnover, or any evidence that a team whose members were familiar with each other had any advantage in close games.

Now, what to make of those findings about hitters? For one, let's first point out that all we have here is an association. Turnover and changing teams is associated with other factors that might be in play here. For example, being a new guy in town means moving to a new home ballpark. One thing I've found previously is that hitters tend to perform better in ballparks that they know well, so perhaps we're just seeing the effect of a batter getting used to a new park. Maybe Albert Pujols (a new player with the Angels last year who struggled to hit HR early in the season) just needed some time to adjust to the new batter’s eye.

Still, the argument that this really is an effect of team chemistry isn't that hard to make, if we assume that low turnover builds friendships (sensible), and that friendships make people happier (very sensible). Science has shown that people show better physical performance when they are happier. Maybe that little extra spring in a player’s step (or swing) is enough to push a ball over the wall that otherwise would have been caught by the center fielder.

But then there's the matter of whether low turnover unto itself is a goal that a team should pursue. Well, if a team has a bunch of Hall of Fame-caliber players, then yes, obviously. Keeping Hall of Famers on your roster is a good thing. But because we're talking about maybe one home run of added value, if a team has a chance to add a player who projects to be five HR better than the player he would replace, it will get more bang for the buck if it simply signs the new player. The effects of player talent level far, far outstrip the effects of low turnover on the roster. It feels so satisfying to think that a group of young players might grow into a World Series winning crew through shared adversity and in the meantime develop great friendships with each other, because that's the plot of every team sports movie ever. But sports movie plots are often terrible guidebooks for running actual teams.

Wow, is it cloudy today…
Let's see if there's anything we can learn from this. If we assume that we really are looking at some effect of team chemistry, how can your favorite team use this information to make itself better? It looks like the worst outcomes are for the guy who is the one new team member in a room full of people who know each other. He's the new kid in school whom your mother always told you to walk over and say hello to. It's good advice for all walks of life.

And really, that's what it can come down to. In my article on how to (theoretically) measure team chemistry, I mentioned that one marker of healthy group dynamics is the presence of social networks that are well-formed. The networks also need to be open to a little bit of change. If there are 24 holdovers from last year and none of them are interested in making new friends, it's going to be a lonely time for the new guy.

But think about what happened when you started a new job or went to a new school or moved to a new town. Society has scripts for meeting new people, introducing yourself, and getting to know someone. I absolutely despise talking about weather of any kind, but I tolerate it, because it's a societally accepted entrepot into getting to know someone. You're not really talking about weather, you're trying to be nice and saying, "I'd like to be friends". Maybe the conversation wanders and you bond. Maybe you end up marrying that person. (For the record, my wife and I did not talk about the weather when we first met. I knew she was the one for me right then.)

A team can help the new guy get acclimated, and maybe hit a few more home runs, by having some sort of simple welcoming ritual. Or having a guy around who is really good at reaching out to people and making them feel welcome. Y'know, a good clubhouse guy. Maybe through that effort, teams can get all the benefits of bringing in a better player without the penalty that he might pay for being the new kid on the block.

Once again, thanks to Ben Lindbergh for the inspiration, and Andrew Koo for the research assistance.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
Would it have been more useful to look at days spent on the active roster, as opposed to appearances in games?
Ideally, yes, but those data aren't easily available, and this is a case of "direction before precision"
Reading this it made me wonder if some organizations are better suited to high roster turn over. Great work because it really made me think even though I think it would be impossible to prove anything because of all the outside factors that are at play here.
Intriguing article, a couple of ideas for follow up:

Are stable teams "luckier", do they pitch better than their peripherals, do the hitters have higher BABIP?

And totally unrelated to performance, Do teams that keep a core group of players do better in attendance. Do the fans find it easier to relate to a team that stays mostly the same from year to year.
Thank you for the follow-up article. The approach covered a lot of concerns I mentioned in my comment on the previous one.

Of the variables you investigated, I didn't see age. It's a reasonable hypothesis that younger players would feel the effects of "chemistry" more than seasoned veterans. This is both because of the sociology (younger players value social interaction with teammates more highly) and because they have lower knowledge/experience (veteran players know what things are important to performance and what can be ignored).

Let's say a team has post-game outings. Younger guys might care a lot about who is invited, who talks to who when they're out, etc. "Chemistry" might matter a lot to those guys. Older players might skip it to spend time with their families and get enough sleep, which might immunize them against potential positive and negative impacts.

Is this easy to tease out from the general trend of "younger players are improving, older players are declining", in terms of performance?
Could there be a selection bias due to age? Would high and low turnover teams tend to have (and add) players of different ages and, therefore, be on different points of the aging curve, biasing your predicted performance based on the preceding year?
At the team level, could you compare pre-season projected winning percentage to winning percentage through the trade deadline? I can see a possible correlation between turnover and in-season roster churn that could mask the correlation with full-season records.
Interesting, but I was surprised you looked at individual performance.

I think the question is do teams with "chemistry" win more games? The non stats folks (Thomas Boswell) would say that teams without "chemistry" have a bunch of individuals chasing individual stats, at the expense of the team.

Perhaps compare pythag vs actual for low turnover vs high turnover teams?
Russell looked at teams, too, comparing pre-season projections to actual records.