Soccer, Basketball, and Where Baseball Lies in Between.

I was recently listening to a podcast called Revisionist History.  Revisionist History, hosted by the (among other things) journalist and writer Malcolm Gladwell, is really, really good, but that isn’t the point here.   In the particular episode I was listening to, titled “My Little Hundred Million”, Gladwell makes reference to two economists who have done work analyzing performances in various team sports.  Those economists, Chris Anderson and David Sally, have written both academic papers and books about their work.  I hadn’t heard of them, but Billy Beane had; he is the first person quoted in the online review of their book, “The Numbers Game: Why Everything You Knew About Soccer Was Wrong”.

In the context of this podcast, Anderson and Sally discuss their juxtaposition of basketball to soccer.  On one hand, their research has found that basketball teams benefit considerably from employing individual superstars.  Soccer teams, on the other hand, historically do best when they employ a greater quantity of good players rather than building around a single player who possesses uniquely exceptional skills.  In short, (whilst the 2018/19 Lakers might offer a decent counter example here…) LeBron James is capable of doing more for his team than Lionel Messi is, due simply to the nature of the sports that they play.

The primary takeaway for their discussion on soccer was this: sooner should a soccer team upgrade its worst player than find a new best player, all else (salaries, etc.) held constant.  As a result, these economists recommendation to soccer owners is clear. Teams should prioritize signing multiple solid players to relatively modest deals opposed to seeking out a single free agent mega-deal.  In other words, in the case of soccer, you’re really only as good as your weakest link.

Here we might imagine that Chris Anderson and David Sally would laud the Dodgers signing of A.J. Pollock and corresponding organizational restraint in talks with the recently-signed Bryce Harper.  Better to sign three “four starters” than a single Ace.

But baseball is not basketball and baseball is also not soccer. Nor were these revelations worth their own paragraph, but alas.

So soccer and basketball lie upon either end of the spectrum.  Naturally one begins to wonder wherein (whereon?) that spectrum baseball lies.  Rather than beginning a search for this answer outwardly where experts and/or economists might readily have an answer, I attempted to answer this question, or a very narrow part of it, by asking my own very narrow question. I wanted to test how a lineups strongest or weakest link might correspond to its overall formidability.  My statistic of choice in this case was weighted runs created plus (wRC+) as it is exclusively an offensive statistic and isolates for various incongruences between teams like park factors.  I collected data from FanGraphs. Very simply, does a teams cumulative average wRC+ correlate more to that any given teams player with the most plate appearances, or to the player who received just the ninth most?

The reasoning essentially being, the player who received the most plate appearances on any given team might not be their best player, but they were clearly good enough to earn a full season hitting atop the lineup.  The player with the ninth most plate appearances might be a fourth outfielder, utility infielder, or a primary pinch hitter in the NL, all of which are generally considered a something like the 21st — 25th spot on a roster.  

Below are two graphs.  The first visualizes how team wRC+ and plate appearance leaders’ wRC+ correspond.  The second visualizes how team wRC+ and “nine-hole” individuals’ wRC+ relate.

The correlations listed as part of these scatter plots make clear that, in the case of 2018, players who received the ninth-most PA’s on their team feature wRC+’s that more closely correspond (0.456) to their teams cumulative average.  As aforementioned, this is a very narrow perspective to take, yet it is interesting that “nine-hole” correlation is rather considerably greater than the PA leaders correlation (0.305).  An, albeit superficial, answer: given these brief data, it appears baseball leans toward the soccer side of the spectrum.

Below are tables featuring the specific players in each scatter plot. These lists are telling.

The table above featuring plate appearance leaders is a mixed bag.  It sheds some light on the weaknesses of this line of reasoning certainly, but highlights some value too.  Average wRC+ in that groups is over 118, well above average as we might expect.  7 of 30 players are actually below 100 wRC+, but many other represent a player who has a relatively strong claim to being his lineups best.

A second Table!

This second fourth table is similar to the third in that it features several real strengths but obvious weaknesses too.  Players like Tony Kemp, Chad Pinder, and Brock Holt respectively highlight the strong offensive depth of the Astros, Athletics, and Red Sox.  Injuries (Justin Turner) and trades (Tommy Pham) illustrate how this dataset is not as representative of weak link/“nine-hole” type players as it is meant to me.

Studying pitching rotations might be another interesting task in this regard, although once again injuries and trades present considerable confounding factors to consider.  Until those factors can be more deliberately addressed, these activities and their brief findings come with a bold asterisk.