How Much to Make of Juan Lagares’ Defense

If you’re not much of a hitter, you might be considered to play a lot anyway for one of two reasons: you’re a pitcher, or you’re an outstanding defender. Juan Lagares isn’t much of a hitter, and he doesn’t project to be much of a hitter, but the evidence and the eyes show he’s an outstanding defender, and that’s why he’s in the mix to start in center field for the Mets. That’s why he’s probably the favorite, or at least, that’s why he probably ought to be. The Mets also happen to be big believers in Eric Young, and that’s their right, and I don’t intend to address that part of the conversation.

What Lagares has on his side are some incredible defensive numbers. We all know to be cautious with those, when we’re talking about individual seasons. The words of this rival team official read like FanGraphs in the newspaper. There’s no question that Lagares is highly talented in the outfield, and that no amount of regression can make him look anything but skilled. But with Lagares in particular, the situation’s a little different, and the numbers have to be treated a little differently.

To quote that official:

As one rival team official noted, defensive metrics generally are plagued with “less year-to-year reliability,” making them far less precise than their offensive counterparts.
[…]
“You can’t just take his 2013 WAR and say that’ll work,” said the official, who has a background in analytics. “You have to regress defensive projections a bit and combine those with your offensive projections, plus whatever else you know about the players.”

Absolutely, right on. Lagares’ WAR last year was 2.9, in a partial season. That doesn’t mean you just project him as a borderline star in a full season. That’s fairly obvious stuff, but Lagares was supported by crazy defensive statistics and those have to be dealt with. What’s really interesting about Lagares is how those statistics break down.

For outfielders, UZR considers errors, range, and arm. With Lagares, you can just ignore the errors column, because the number is negligible. His range number was terrific, probably because his range was terrific, and this is what people usually think about first when they think about defensive statistics. They think of them as measures of range. But Lagares’ UZR got an enormous boost from his arm. His arm rating was more strongly positive than his range rating.

We have arm ratings going back to 2002. Let’s put everything over a per-150-games denominator. Last season, per 150 games, Lagares’ arm comes out to 20.6 runs above average. Since 2002, there have been 922 player-seasons of at least 750 innings in the outfield. Lagares in 2013 owns the highest Arm/150, by a full four runs over 2007 Alfonso Soriano. Right behind, there’s 2004 Alex Rios. Over 12 years of baseball, at least, no outfielder had a more productive arm than Juan Lagares last summer. He wound up with an astonishing 15 assists.

So, yeah. But then, if you paid any attention to the Mets, you already knew about this. It was no secret that Lagares was doing things with his arm other players generally didn’t do. It was actually a function of Lagares’ aggressive positioning, speed, hands, and arm. Controlling the running game is a skill of Juan Lagares’, but more important than what he’s done, now, is what he’s going to do. How reliable is an arm rating, year to year?

My guess was that range rating would prove more consistent than arm rating. I identified all outfielders who played at least 750 innings in consecutive seasons between 2002-2013. This left me with a sample of 587, and then I calculated all the Range/150 numbers and all the Arm/150 numbers. Year-to-year, Range/150 yielded an r value of 0.50. Here’s what that means: if a player was one standard deviation better than average in Year X, he’d be expected to be 0.50 standard deviations better than average in Year X+1.

Year-to-year, Arm/150 yielded an r value of 0.32. As expected, there’s a weaker relationship, and here’s what that looks like graphically:

You might prefer a table. Here’s another way of showing what regression looks like. For Year X, I split the players into six groups, in descending order of Arm/150. Then I averaged their Arm/150 ratings for Year X+1.

Group	Year X	Year X+1
Group 1	7.7	2.4
Group 2	2.9	0.6
Group 3	0.8	-0.2
Group 4	-0.9	0.1
Group 5	-2.9	-1.3
Group 6	-6.0	-2.2

The idea was already that we had to regress Juan Lagares’ defensive numbers. But because he built so much value on his arm, we have to regress a little more aggressively, because the arm numbers are a little less consistent than the range numbers. And that seems to make intuitive sense, because there are fewer opportunities to make a difference with your arm, and once people know you’re a threat in the outfield, they’re likely to change their baserunning strategies.

Just for the sake of showing some individual cases, everybody agrees that Jeff Francoeur has an amazing arm. In 2007, his arm was worth 16.6 runs above average. In 2006, it was +3.5, and in 2008, it was +2.5. Ichiro went from negative to positive between 2002-2003, and he went from positive to negative between 2008 and 2009. Alfonso Soriano lost ten runs per 150 after his remarkable 2007. Nyjer Morgan lost 20 after 2009. No single player has yet posted consecutive double-digit Arm/150 ratings. Well, let me revise that — no single player has yet posted consecutive double-digit Arm/150 ratings with the same sign. Between 2002-2003, Bobby Higginson went from one of the best ratings to one of the worst. Jeff Francoeur has almost pulled it off, but Jeff Francoeur has almost pulled a lot of things off.

Getting back to Lagares, last year his Arm/150 was almost 21. It was about 4.4 standard deviations above the average. According to the calculated r value, next year we’d expect Lagares to be about 1.4 standard deviations above the average, which would leave an Arm/150 around 6. Which would represent a reduction of about a win and a half, and truth be told we should probably be regressing more heavily since Lagares was in the outfield for barely 900 innings. As noted before, you can’t make Lagares’ arm not good. But you can make it less extraordinary, and applying regression makes Lagares’ numbers a lot more realistic.

You regress the arm stuff, heavily. You regress the range rating separately, and then you put them together. You still get a very good defensive center fielder, and maybe Lagares is better than his regressed self, but you can begin to understand why the Mets aren’t just going to hand him a regular job. His 2013 WAR is misleading. Lagares really might not be able to hit enough, because as good as he’s been in the field, he’s been that good for two-thirds of one season.

Sure does still seem better than Eric Young, though. And I sure do hope people keep trying to run on him, because as much as I believe in regression to the mean, I’m most fascinated by those who refuse to do it, and we can’t be sure it’ll happen to Juan Lagares until or unless it happens.