06 March 2015

The Misunderstood Relationship Between Team Chemistry and Sabermetrics

Almost two weeks ago (I was hoping to write about this sooner, but life gets in the way sometimes), Steve Melewski of MASN ruffled some feathers with the following tweet.

Overall it seemed like an odd statement to make.  Not odd in the sense that some of the major league pitchers decided to watch Hunter Harvey throw his bullpen session, but odd that he appeared to force the topic of team chemistry and sabermetrics into the observation.  It’s entirely likely that the inclusion was only meant to get a reaction, which it did.

The fact is he’s not incorrect; sabermetrics can’t put a stat on team chemistry*.  However, his tweet implies that sabermetrics completely discounts team chemistry, which is simply not true.  The way that team chemistry relates to production on the field has been studied quite a bit, but it just so happens that it is incredibly difficult to quantify its effect.  So while you can’t put a sabermetric stat on team chemistry, it hasn’t been for a lack of trying.

*He’s also kind of incorrect, as people have actually tried to put a number on team chemistry, although that attempt is not without its own issues.

In particular, Russell Carleton of Baseball Prospectus has tried several times to quantify the impact of chemistry on a team’s production.  He’s examined Brandon McCarthy’s claim that the 2012 Oakland A’s (who won 94 games) would have been a 70 win team without the presence of Brandon Inge and Jonny Gomes** (possibly, but not nearly to the extent McCarthy claimed), whether having a veteran with a good clubhouse presence helps a young player produce better (it doesn’t), and whether players perform better on teams with high personnel turnover or low turnover (some correlation for hitters, but nothing for pitchers, teams, or a team’s performance in close games).

**Inge and Gomes combined for 2.8 WARP in 2012, according to Baseball Prospectus

While those results initially tell us that the effect of team chemistry is minimal or even non-existent, Carleton himself points out that it doesn’t mean that there isn’t some real positive effect of mentoring a young player.  In fact, in one of his non-gory math articles about chemistry, he wonders what “chemistry” actually means, eventually hypothesizing that it may boil down to whether an individual cares enough about his team (and teammates) to not take certain plays off.  However, as Carleton states, that decision to not take certain plays off could also be made out of pure selfishness (for example, playing hard to get a higher paycheck the following year).

Differing opinions about team chemistry aren’t limited to sabermatricians and beat writers, they vary between people within the game itself.  On one hand, there is Los Angeles Dodgers starting pitcher Brandon McCarthy (who has a reputation as a believer in sabermetrics), stating his belief in the unaccounted value provided by Inge and Gomes (mentioned above).  On the opposite end of the spectrum, you have former manager Jim Leyland saying that team chemistry isn’t worth a “hill of beans” (Leyland didn’t actually use the term “hill of beans”, but this is a family website).  In the middle, there’s front office man Jonah Hill Paul DePodesta, who doesn’t doubt its importance, but questions a team’s ability to manufacture it.

The truth is it can be easy to mock the “team chemistry” argument.  It’s only typically talked about before and after the season.  In spring training, EVERYONE has good chemistry.  Ben Lindbergh wrote about that for Baseball Prospectus last year and found that 25 of the 30 MLB teams believed they had great clubhouse chemistry prior to the 2013 season (spoiler alert, not all 25 teams won the world series that year, or even managed a winning record). After the season ends, as Lindbergh states, “chemistry is trotted out to explain success or failure after the fact.

However, that doesn’t mean team chemistry is irrelevant.  The difficulty is in measuring its effect, especially without everyday access to record the complex interactions that 25 individuals have on a daily basis over the course of a long season.  The ones with the closest type of access are beat writers.  Still, they don’t see everything, and wouldn’t necessarily report on it even if they did, because they require a certain level of trust with the players in order to effectively perform their job.

Melewski’s statement on team chemistry and sabertmetrics wasn’t technically incorrect.  However, it was unnecessary and stated in a manner that encouraged confrontation.  Baseball teams are made up of many unique individuals with different personalities and motives.  Can we be sure that Tillman or Britton weren’t watching Harvey’s bullpen session out concern for their own job security?  Maybe they decided to observe out of sheer curiosity?  I mean, who wouldn’t want to watch a top 30 prospect pitch?  Those may sound like silly questions, but I don’t think they can be completely ruled out.  Team chemistry is an extremely complex concept that has yet to be quantified (and quite possibly it never will).  Watching a teammate throw a bullpen session shouldn’t be touted as definitive evidence that it exists.


Note: While this article mostly used examples from Baseball Prospectus, other websites (such as Fangraphs) have previously looked at the topic as well.  Additionally, past work by Camden Depot’s Jon Shepherd on manager wins added (or subtracted) could be considered an indirect way to measure team chemistry.


Northern Oriole said...

Sometimes articles about baseball analytics are insightful and stimulating, most often just dull, but sometimes border on self-parody. Maybe it's best to focus on improving something like measuring defence before moving on to the more metaphysical aspects. The questions these debates always raise for me are: why is the ananlytics crowd seem so sensitive? Why do they take things so seriously? And what does any of this have to do with the joy of baseball itself? Anyway, well-written but I don't think it's a fruitful endeavor.

Jon Shepherd said...

People tend to like to control how they themselves are defined instead of being disparaged. We are loud and we are proud.

Anonymous said...

"Not everything that matters can be measured and not everything that is measured matters."

Jon Shepherd said...

Also, second note, psychosocial aspects impacting group performance is something that has been a rather strong field for decades. It makes sense for other organizations outside of the normal working world to try to understand how personal connections impact performance on playing fields. This is why we have seen considerable efforts in basketball and soccer the past few years trying to add some clarity. I imagine that baseball, football, and other team sports have also dabbled in this.

I mean, why not? It does not hurt to look and see if there is some low hanging fruit to pick.

Jon Shepherd said...

Anon - the quote is "It would be nice if all of the data which sociologists require could be enumerated because then we could run them through IBM machines and draw charts as the economists do. However, not everything that can be counted counts, and not everything that counts can be counted." Your quote is the often misattributed misquote directed at Einstein.

However, this is not what is contested. Trying to find impact does not mean proving there is no impact. The key point is providing effort to determine if something can in fact be counted. There was certainly a time when an atom was thought impossible to know. And, knowing the presence of an atom, does not extinguish the beauty that is life. Similar things would be true if someone could figure out any aspect of psychosocial aspects impacting play in baseball.

Philip said...

I saw nothing controversial about Steve's tweet. It seemed to be a joke and nothing more.
Having said that, people need to accept that just because something can't be measured has no bearing on whether it exists or is important.
The best example, God, is also the oldest example.
Chemistry is very real and very important, and beneficial chemistry is better than the negative version...even if it can't be measured.

Philip said...

BTW, very interesting article. I hope you will return to this subject frequently.
It is extremely interesting to watch the development of detailed stats as more information is collected and analyzed.
I'm especially interested in defensive stat progress. Defense isn't well represented in the Stat world, which is one reason the Orioles have been discounted over the past few seasons.

Jon Shepherd said...

In conversations with Steve on Twitter and secondhand ones, it was indeed a joke, but he was also making a point that is actually not well versed in what data science is actually trying to do.

Matt Kremnitzer said...

I took it as a troll-ish comment; it's also not the first time he's written something similar.

Philip said...

Jon, are you going to resuming writing for this site?

Jon Shepherd said...

No, I have a meandering article that probably will be posted by BP soon on Clayton Kershaw and Mike Pelfrey.

My time at the Depot was nice, but my contributions these days are typically behind the scenes and the rare comment.

Anonymous said...

I just wish they would watch Ubaldo Jimenez throw...