Alex Gordon and the Tale of fWAR

Kansas City Royals left fielder Alex Gordon and wins above replacement (WAR), notably fWAR (Fangraphs WAR), made news this week, as Gordon at one point topped the WAR leaderboard. As of this moment, Gordon remains of known superstars, such as Mike Trout, Giancarlo Stanton, and Robinson Cano among others. I am sure most have heard of Alex Gordon at the least, if not have seen play, knowing he is a quality baseball player in all aspects, but is he really deserving of the lead in fWAR? Is he the best player in baseball?

In general, WAR attempts to encapsulate everything a player does on the field (offense, fielding, base running, position, etc.) and assigns an objective, statistical value that gives, in theory, the number of wins said player has over a replacement player. What is a replacement player? In theory it is an individual with 0.0 WAR, but in practice this is not the case, as names thrown about as replacement players often have more than 0.0 WAR on average a season. A number of people consider AAAA guys replacements level. AAAA essentially means non-prospects that perform fine in AAA, but are too mediocre to be an everyday Major League Baseball player, to say the least. This is more of an indictment of calling the WAR measurement just that, wins above a replacement level player. If a replacement level player is above 0.0 WAR on average, then how are they a replacement player? You may see the confusion the statistic already brings about. If the baseball community were to continue on with WAR in any form, it should attempt to stay true to its objectives and title, wins above replacement, not wins above Jason Kipnis-level. At this time, there are no players that sit at 0.0 on Fangraphs WAR leaderboard (although there are negative WAR players). Even in practice, if team A’s catcher were to go down for the year, the replacement catcher that is either signed, called up, or traded for is likely not to produce a 0.0 WAR, which, again, makes the basic statistic misleading on a fairly fundamental level.

Specifically, how is WAR, or in this case fWAR (Fangraphs WAR), calculated? Well, that is one of the issues with fWAR, unlike other statistics, is that it is not transparent in the slightest. Those outside of Fangraphs have no idea how it truly is calculated. There is no formula for fWAR. There is a mathematical formula for VORP (Value Over Replacement Player, similar to WAR) or wOBA (weighted on base average, which looks solely at offense production). Fangraphs lists the components for position players’ fWAR (Weighted Runs Above Average, Ultimate Zone Rating (rSB and RPP for catchers, do not worry too much about this), Ultimate Base Running, and Weighted Stolen Base Runs, along with park factors, positional adjustment, and replacement level (Do not worry too heavily about these things)), but does not give a formula assigning weights and how exactly these numbers interact to produce Fangraphs Wins Above Replacement. This lack of transparency is concerning. How are these numbers calculated? We do not know, we just have to trust Fangraphs. In just basic math, if one were to work a math problem and produce a value that looks unlike the assumed result, they are able to go over the problem, perhaps reverse engineering the problem to see what step may have produced an erroneous value, but we cannot do that with fWAR and that is a serious problem.

WAR, notably fWAR, is used for both pitchers and position players, and although pitching WAR is one of the worst ‘advanced’ metrics in baseball, this is about the Alex Gordon situation, so let us focus on position player WAR. My other major qualm, alongside the lack of transparency, is defensive metrics, not only on Fangraphs, but defensive metrics in general. Although I say that, the first value to go after is Fangraphs value of defense, which looks at both fielding and position adjustment. Again, this is Fangraphs defense rating, not to be confused with anything else. Alex Gordon ranks third overall using Fangraphs defense rating, at 17, which means Gordon has saved 17 runs more than the average player. This rating takes into account positional adjustment. There are two points to be made regarding this statistic; how is it calculated and how in the world can someone in left field be that much more valuable than say an elite defensive shortstop after taking into consideration positional adjustment? The answer to how it is calculated is an even bigger mystery than fWAR. Fangraphs does not even bother to attempt to list the factors that make up this defensive rating. We do not know how it is weighted and we do not have a hint as to what defensive metrics are used to determine the rating? How can one possibly justify this defense rating of having any value of the slightest without even the slightest hint of transparency? As for the question regarding left field defense being more valuable than elite shortstop defense, again, that cannot be answered because we cannot determine what makes up this rating, so the answer is impossible unless you are of access to the formula.

Comedically, this mystery makes up a number of widely used defensive metrics, namely UZR (Ultimate Zone Rating). UZR attempts to look at runs saved as most other advanced defensive metrics attempt to do. Fangraphs calls UZR “publically available”, which is absolutely false. The ratings themselves are available, which they do not specify what they are referring to when they say publically available, so it may be that, but as far as the method and formula for calculating UZR, it is another mystery. Again, it boils down to trusting the creators of these privately designed formulas, which is an obvious issue. Another major issue is that UZR is essentially a waste, as UZR does not factor in plays where a shift is used. The shift has been used on 8,134 plays where the ball was put into play from 2011-2013, a fairly significant amount of plays to just toss aside. UZR also sees fairly large swings from season to season. For example, in 2013, Andrelton Simmons, an elite shortstop, posted a UZR of 24.6. Thus far in 2014, his UZR is 8.0, nowhere near on pace to reach 24.6 or even half of that 2013 value. Did Simmons become less than half the player defensively this year in comparison to 2014? While you can make the argument that Simmons has not been quite as good this year, such a large drop-off has to raise some eyebrows. Simmons is not an isolated incident, but the list of players to check out would go on and on, so just check out some quality defensive players’ UZR ratings on Fangraphs through various years. The problem with UZR is not just UZR itself, but the fact that other compilations statistics, such as fWAR, include UZR to get a calculation. How can the compilation statistic be valid if the underlying makeup (UZR) is full of too many holes or questions to be considered valid itself?

Basic defensive metrics are largely inaccurate or fairly subjective. Errors are scored not by an objective individual in each case, as it is up to the official scorer, who may be incompetent or biased at times. Darwin Barney’s 2012 season, in which only two errors were attributed to Barney, had dubious validity, as there were multiple occasions in which Barney was not charged with an error when he should have been. Al Yellen takes a look at one occasion specifically against the Pittsburgh Pirates, where Barney gave up an error, but it was not scored as one. These sorts of incidents are far from rare. With such weak validity, how can errors be used objectively? How can you assign any value to errors? Double plays are almost always infielder-based statistics and how much value do they have anyway? The goal of defensive metrics is to assign value of some objective values to a player or team, but it just is not there at this time. At this point in time, the eye is the best measuring tool of defensive prowess. Of course, the eye is far from objective, as people’s opinions vary from time to time and person to person, but the statistics do not hold up, while the eye does.

As an example, let us look at the individual who has the spotlight on him, Alex Gordon. Let us go back to that Fangraphs defense rating. Alex Gordon sits, again, at 17 runs saved defensively using said metric, which places him at number three overall. Alex Gordon is a really good left fielder, almost certainly the top left fielder defensively in baseball. Mike Trout, an elite center fielder and among the discussion as the top center fielder in baseball (I am sure some will still argue Bourjos or someone else), gets a defense rating of -7.1. Again, how are these ratings calculated and what factors are used is absolutely unknown, so we are just going off of the word of Fangraphs. Is anyone seriously under the belief that Alex Gordon saves 24 more runs than Mike Trout? I would not buy that belief that Gordon saves any more runs than Trout. This is not just an indictment on Gordon, but really of the entire metric. Check out the list to see various absurdities (Brian McCann saves more runs than Jonathan Lucroy?). UZR is hardly better, giving Gordon a rating of 24.8, which is a rating of essentially an all-time great season defensively. Again, Gordon is really good, but these numbers help shed light on a problem that has been plaguing baseball for a number of years: WAR and defensive metrics.

At this point in time, defense is so hard to measure objectively. Major League Baseball Advanced Media and Sportsvision’s field f/x appears to be fairly promising, as it seems to measure path efficiency along with play difficulty fairly accurately and objectively, but at this time, the data is not available and the technology used to track the players is only installed in a limited number of parks. It is unknown whether this technology will ever be made public considering the size of the data files (several terabytes for data for one game, if I recall correctly). Field f/x is not the complete answer to all of one’s fielding inquiries, but it is certainly more accurate than the defensive metrics currently being employed.

Going back to our initial point of Alex Gordon leading the board among all players in fWAR, hopefully you have some understanding as to why this metric and judgment is meaningless. In defense of the metric, some have attempted to shift goalposts, stating that WAR is not an indicator of the best player in baseball, just a measurement, but in theory it does determine who the most valuable player is, as they attempt to measure the value of a player, even adjusting to position. After the positional adjustment, I am not quite sure how you can attempt to say it is anything else. Even if one were to somehow hand wave that argument away, fans and writers alike misuse the statistic (both because it is fundamentally flawed and because it is not supposed to be used to determine who the best is), making the statistic an even larger joke. As far as baseball fans go, as a whole, we have gotten lazy with the development and the usage of WAR. It is such a poor statistic that encompasses all, or at least attempts to, thus leading to certain individuals citing WAR and then just moving on. I do not use WAR in any fashion in articles, as it is just a waste of time and makes work appear questionable.

The Alex Gordon: American League Most Valuable Player articles are almost embarrassing to read, with some going as far as to state Gordon as the best defensive player in baseball. In the hierarchy of positional value, left field is towards the bottom of the list, as it does not require the arm of a right fielder, the athleticism of a center fielder, nor the quick skill of many infielders. It is not a hard position to play, relatively speaking. Shortstop is a hard position to play. Center field is a hard position to play. Catcher is a hard position to play, even with my qualms of pitch framing value.

Let us look at what we know, which is offensive statistics. Those are not shrouded in mystery, but have formulas out in the open and are fairly reliable and accurate. Offensively, Gordon is the best player on the Kansas City Royals, which is not a monumental achievement given their offense. Gordon has an OPS of .810, but to be more specific let us look at both weighted on base average (wOBA) and Weighted Runs Created + (wRC+), both of which are fairly accurate and transparent ways to measure an individual’s offense. Gordon has a wOBA of .357 on the year, which is pretty good, but nowhere near MVP caliber-offense, especially this year. Gordon has a wRC+ of 128 on the year. Think of wRC+ as a more accurate OPS+, where 100 is league average. A wRC+ of 128 is good, perhaps All-Star caliber if that means anything, but is just that good, not great, not amazing, and certainly not MVP-level. Just going down the wRC+ leaderboard in descending order, one would find Gordon at #39 in all of baseball. Even being friendly and looking at just the American League, Gordon ranks #21. These are accurate ways to measure offense, so by going off of offense alone, one would get quite the chuckle of a 128 wRC+ player getting MVP discussion. That is the known in this case. The defense of Alex Gordon is the known and the unknown. What we do know is that Alex Gordon plays really good defense, perhaps great defense, although it is at left field, which is not a valuable position, relatively-speaking. What does being the best left fielder in baseball place you? It is hard to say, but the top fourth maybe? I might be somewhat generous when you consider the difficulty of the other positions on the field, so it might just be in the top half of baseball. I made the remark earlier that you would have to literally be Superman in left field to be the best defensive player in baseball (perhaps the Flash is more fitting).

Where is Gordon in the American League MVP discussion? I would say top ten for sure and would argue top five. There is the real American League MVP, Mike Trout, ahead of him, as Trout is better by a mile offensively (note: the mile measurement is metaphorical, not a real baseball measurement), is the better base runner, plays better defense at a more valuable position, has a better arm. Others in the discussion to be the runner-up to Trout include Jose Abreu, who plays a lesser position at first base and is average at that, but is otherworldly on offense, arguably better than Trout with the bat, Victor Martinez, who plays first base or is a designated hitter, but is, like Abreu, a great player with his bat, Jose Bautista, who is more productive with his bat and plays above-average defense in center field and right field, Michael Brantley, who is better offensively and plays both left field and center field fairly well, and Miguel Cabrera, who offers little at first and third base, but is considerably better at offense. I am not necessarily stating that those individuals are without a doubt better (this is not including Trout, who is, without a doubt, better), just people in the discussion alongside Gordon in the American League Most Valuable Player race.

The ultimate takeaway from this should not just be that Gordon is not the MVP. That is not even the primary lesson. The primary lesson is that you should do research into metrics before using them. Unfortunately, far too many people discuss WAR and various defensive metrics like they have some sort of meaning, when I have outlined the fundamental flaws with them. Do not be lazy.