On Monday, I wrote about how the loss of Dexter Fowler did not make or break the season. In that article, I wrote about defense in a very conventional wisdom-y sort of way:
The above obviously does not take into account fielding prowess of any of the players. [Nolan Reimold and the rest of the current COFs on the roster are] defensively worth about -10 runs, which would be the same as the David Murphy, Pedro Alvarez, David Freese, and Matt Kemp options. At -5, Dexter Fowler, Jay Bruce, and Nick Markakis. Austin Jackson and Oswaldo Arcia fall in either at league average or +5. Marcel Ozuna is either at +5 or +10. Anyway, that is the rough way to look at it through this one data science way of looking at it.And that is a rough way to look at it. In fact, the whole center to corner conversion has been gnawing at me for a few months because we simply wave a magic wand and simply say +10 runs. To test this all out, I looked for centerfielders who shifted over to right field in their late 20s and early to mid 30s. I had two approaches, a simple conversion and then regression modeling. I created a regression model which compared their last two seasons at centerfield to their first two seasons in right field. I included age modeling, but by this point in a career, fielding is largely settled with only minor shift. Your major decreases in fielding tend to occur in a player's early to mid twenties. Anyway, I tested age as a variable and found it meaningful only for error rates, which is the only metric (i.e., UZR's ARM, RangeR, Err) that appears to greatly change within this age group.
The players considered in this sample were likely quite few (10): Carlos Beltran, Ichiro Suzuki, Torii Hunter, Curtis Granderson, Randy Winn, Shane Victorino, Marlon Byrd, Alex Rios, David Dejesus, and Cody Ross. With so few, there certainly might be some selection bias. As in, the players in this grouping might be unique in comparison to the players we might choose to compare this group to. Austin Jackson and Dexter Fowler might be different from this grouping in some way that would meaningfully make any comparison have minimal value. That said, I cannot imagine at the moment why.
Anyway, there are certainly differences between defensive values from the centerfield values vs. the right field values (0.01) for this group. A simple averaged conversion would be to adjust the ARM, RangeR, and Err values by 1.5, 7.6, and 0.1, respectively, over 150 games. That, of course, sums up as 9.2 runs improvement. That gives some credibility to the nonchalant +10 mark.
Here are those results:
However, the regression model (0.17) sees things a little different.
I added Adam Jones in there just to see how he is projected with a move to right field. Marcel Ozuna I considered because the Miami Marlins are looking to replace Carter Capps. It probably is a pipe dream, but the Marlins could deal Ozuna and then sign Jackson. Anyway, I simply wanted to put that projection on the table. The real comparison I was interested in was the one between Jackson and Fowler.
As you may remember, the lineup modeling tool projected Fowler's presence in the lineup to be 13 runs greater than Austin Jackson's contributions there. However, the defensive modeling performed in this post put Jackson's glove trumping Fowler by 10.4 runs. In the end, that suggests Fowler is worth only 2.6 runs more than Jackson. That is about 0.3 wins. The regression model is indicating that a blanket 9.2 run increase in defensive value might well be inappropriate.
If we consider the regression modeling problematic due to having low significance (0.17), then we might feel more confidant going back to the original averaged conversion. In that case, our expectation is that Jackson is 13.7 runs more valuable than Fowler, defensively. All together, it makes Jackson 0.7 runs more valuable.
So, yes, as I wrote on Monday, keep calm. Fowler probably was not that big of a catch.