Alex Gordon is one of the top candidates for the American League MVP, maybe. He leads the league’s hitters in WAR, as measured by FanGraphs, and even if that’s not the only way to decide who should be the MVP, it’s a pretty good one. Alex Gordon: MVP conversation subject. Get used to it. You’re going to hear it a bunch. (Dave Cameron makes many excellent points about this in a piece at FanGraphs today.)
We all know, or should know, that defensive metrics are less reliable than offensive metrics, particularly over the course of a single season. While offensive stats can fluctuate because the player’s performance itself fluctuates over smallish samples (health, swing mechanics, hot/cold streaks, just had a new baby, whatever), defensive stats fluctuate for the same reasons--and also because (depending on the metric) they can be skewed by opportunities (how many chances does the fielder get, and what sorts of chances are they) and by each metric’s imprecise or subjective assessments. We all know this. There’s a margin of error, and it’s bigger for defensive metrics than for offensive.
Of course, the margin of error applies both ways. So while you’re arguing for or against an MVP candidate, you might say some player’s +5 rating on some defensive scale overrates him. Or you might say it underrates him. We can’t really know. Maybe Gordon isn’t a +19 defender, as UZR (the defensive component of FanGraphs’ WAR) says* he is. Maybe he’s a +29 defender. It’s just as likely to underrate him as overrate him, right?
Actually, no. I’ve got a defensive metrics lifehack for you. If you want to find the guys who a metric is overrating, simply run your finger up the leaderboard until you get to the top. That guy? He’s probably being overrated.
This will almost always be the case when you’re talking about a stat that has any sort of “luck” component to it. To name an obvious example: the league’s saves leader always leads the league in saves opportunities. He usually pitches very well, too. But he always gets some benefit from luck.
To name an example that’s more similar to the defensive metrics: We know, more or less, how to isolate the luck portion of ERA, by calculating FIP. Since 1994, 40 pitchers have led the league in ERA, and 33 of the 40 have had a FIP that was worse than their ERA. These pitchers all pitched very well, and if you wanted to find a pitcher who had pitched well, shoot, looking at the top of the ERA leaderboard would get you one. But if you wanted to find a pitcher who had benefited specifically from luck, the top of the ERA leaderboard would get you one.
Same for defensive metrics. Just look at the ratings of those guys who have led the league, since 2005, in FRAA, Baseball Prospectus’ defensive stat. Typically, they have been +20 to +30 defenders the year they led the league. The next year, they dropped. This is expected, not at all damning; the league leader in almost anything will drop the next year, because you usually lead the league by having a “peak” season. The league leaders in offensive WAR (Baseball-Reference’s) over the same time period, for instance, usually dropped, too.
But just admire the size of the drops. On average, the offensive leaders--let’s say the top two, actually, for each year--saw a 29 percent decrease in the same stat the following year. They were great hitters who had career years, roughly 30 percent more productive than usual.
Now the defenders: The defensive leaders (top two again) saw a 70 percent decrease the following year. They were good fielders who had career years and also probably got overrated some.
So, if you’re looking at Gordon, you can safely adjust his WAR downward by exactly-- hey, what are you doing, dummy? We’re not doing anything exact here. That’s the point! This is noisy stuff, and leaning on these decimal point differences as though they’re conclusive is silly. What makes WAR so beautiful is that it gives you a good guide and a good model for how to think things through, but it also leaves you room to adjust in your mind. In my mind, I’m pretty confident that Alex Gordon’s having an absolutely fantastic season, and also that he hasn’t been quite as valuable on defense as the numbers say.
* Okay, when I wrote this Monday, Gordon was at +19 defensive runs, per FanGraphs. On Tuesday he’s at +17 and now ranks fifth.
