Sunday, September 14, 2014

Thoughts on Player Value Part III

In Part I of this series, we discussed some basic types of baseball data and statistics.  In Part II, we looked at defensive metric systems, specifically UZR and how it is derived.  Bill James, a pioneer in statistical analysis of baseball, or saber metrics, wanted a way to translate player performance into value for his team.  Since the goal of all this is to win games, he wanted to quantify how much each component of what we can measure about a player's performance contributes to his team winning.   He came  up with a number he called Win Shares.

To put it in simple terms, when a player crosses home plate and scores a run, some of that was produced by that player getting on base in the first place and some was produced by the player who got the hit to drive him in.  Even if the player hit a HR, which on the surface would appear to be 100% the work of that player, what about the player who might have gotten a hit earlier in the game who neither scored a run nor drove one in, but his hit allowed the guy who hit the HR that one extra AB late in the game?

As time went on, statisticians were able to tease out averages of how many runs a walk, single, double, triple, HR, SB would produce and how many potential runs would not score as a result of a strikeout or CS.  Since different ballparks make it easier or harder to produce runs, they came up with adjustments for Park Effects.  These parameters were then all fed into a very long equation and a single number was produced that has become known as Wins Above Replacement or WAR.  Replacement is defined as the number of wins a team made up of "replacement players" or players brought up from AAA in an emergency would theoretically win, based on historical numbers.

When you look at lists of players ranked by WAR, you find names near the top that you do not find in more lists ranked by more traditional stats.  Since WAR is a single number that theoretically tells you the global value of a player, the number has become a point of intense debate when end-of-season discussions about who should be named MVP come up.  If WAR is, in fact, an accurate measure of global player value, then it should be simple to identify the MVP.  It's the guy with the highest WAR!

There's the rub.  Just how accurate is WAR for a given player?  The derivation of the number contains so many data points, so much subjective judgement(particularly on the defensive side) and so many adjustments for things like Park Factors, the potential for a harmonic convergence of error to produce an erroneous outlier would appear to be relatively large.  The WAR apologists can point to cumulative WAR scores that closely match numbers of team wins as a validation of the statistic.  WAR critics point to surprising names at the top of WAR rankings as reasons to question whether it is always accurate for individual players.

One such name that comes up over and over is Ben Zobrist who by WAR was the second most valuable player in all of baseball from 2009-2012, trailing only Miguel Cabrera and by just a small margin.  Zobrist is a good player who does a lot of things well, but who does not do any one thing great.  Is he really almost as valuable as Miguel Cabrera who was crazy good during those years?
Since WAR is the only way to objectively measure that, the argument quickly becomes circular.  Neil Weinberg of Fangraphs wrote a recent piece in the Hardball Times section arguing that the reason why Zobrist is so valuable is because of his position versatility, except that WAR theoretically levels the playing field for different positions.

Right now, Alex Gordon is tied for 3'rd place in MLB WAR at 6.0 despite a pedestrian OPS of .792 and playing LF to boot.  Is Alex Gordon really the 3'rd most valuable position player in all of baseball? You really have to wonder, especially when you see that his defensive number is almost 4 times as high as his average for the past 3 seasons and higher than his past 3 seasons combined!

In 2012, Martin Prado, a versatile player similar to Ben Zobrist, put up a WAR of 5.5 for the Atlanta Braves, fueled in part by a similar defensive outlier.  This may have contributed to Dave Cameron's assessment of the trade for Justin Upton as being a "win-win."  As in the Prado-Upton trade, I personally do not believe that a straight up trade of Miguel Cabrera for Ben Zobrist in 2009 would have been a "win-win", and the team receiving Zobrist would not be the one with the win!

So, should WAR be the determiner of MVP?  My answer is it should definitely be a strong contributing factor in the decisionmaking process, but you cannot blindly give the award to the guy with the top WAR number.  You need to be aware of the limitations of the statistic and the potential for it to deliver outlier numbers.


  1. Agreed. But again, there are two measurements of WAR, from Fangraphs and Baseball Reference, with somewhat differing emphases. There are also other measures of the components of WAR, such as Baseball Prospectus' Value over Replacement. Using these various systems to complement each other introduces more subjectivity, in deciding how to weight each in one's final analysis, but also permits the biases built into individual systems to be evened out. And of course any of these systems improves vastly on amateur over-the-cracker-barrel opining.

    1. Not to disagree with your overall point, but the idea that the biases built into the different systems "even out if you average them is itself an assumption.

    2. Yes. By "even out" I didn't mean that they achieve objective truth, if in fact there is such a thing with our current state of knowledge and the divergent sources of value. I meant that diversifying what you rely on is the best way to lessen the distortion that idiosyncrasies produce, and that in baseball evaluations one can easily diversify.