Joe Kelly and why the Red Sox should not add another starter

As the Winter meetings came and went, the Sox now have 5 arms with MLB starting experience to fill out their rotation. However, as fans and the media continue to suggest, we do not have an "ace" or "top of the rotation arm" at our disposal. While I don't want to get into the meaningless debate of what constitutes an "ace," I would like to show you why the Red Sox should rest easy with the rotation we have and not make the mistake of committing nine figures to James Shields or Max Scherzer, nor trade away valuable long-term assets for another 1 year rental.

Based on that paragraph alone, I would expect many of you to disagree with me, and some will likely still disagree with me even after finishing this post. But my goal is to make a compelling case that while upgrading from Joe Kelly to one of Hamels, Zimmermann, Cueto, Shields, or Scherzer would increase our chances of winning in 2015, the marginal upgrade is not as big of a marginal upgrade as what meets the eye. All of the stats and projections I am about to reference are available at Fangraphs.

So without further ado, let's take a look at our current projected starting rotation. We are currently projected by Steamer to have the 6th best starting rotation in baseball at 10.8 WAR (behind the Cardinals, Mariners, Indians, Nationals, and Dodgers). Given that this same projection system believes our lineup will produce 2 more WAR than any other team, and thus to have the best record in baseball (and that's including a projection of Mookie Betts to only get 310 PA and Rusney Castillo being sub-replacement level, which little evidence suggests will be the case). That's not a bad place to be by any means, and I don't think it's crazy to think that as constructed, the Red Sox can win 90+ games.

That last paragraph is mostly for the sake of context. Of course, it would be better to project to win 93 games than 90, to add in insurance for injury or underperformance. It would thus appear that adding one more elite starting pitcher would be the most effective way to do so. Let's have a look at our starting rotation's projections in terms of WAR:

Porcello 3.0

Buchholz 2.1

Masterson 1.9

Miley 2.0

Kelly 1.1

I'm not here with the goal of validating or invalidating Steamer as a projection system, but none of these seem too outlandish. By these metrics, it makes plenty of sense to upgrade Joe Kelly's spot in the rotation. However, I'd like to take a closer look at Kelly to see if there's more than what meets the eye. Steamer uses a FIP-dependent projection for WAR, but there's evidence to believe that Joe Kelly's unique profile as a hard-throwing groundballer with 3 above average pitches would lead to RA-9 WAR being a more effective tool in determining his worth.

In terms of arsenal, Joe Kelly throws 4 pitches, a fastball, a curve, a slider, and a changeup, all of which grade out as above average by Fangraphs except for his slider, which could either be small-sample-size noise (he only throws it 6.7% of the time) or a sign that he should probably not throw it at all. By quantitative measures, his stuff is well above average. If he had enough starts to qualify, his fastball velocity would rank him 7th in the league among starters, right below Stephen Strasburg. All three of his effective pitches are in the upper third of MLB in terms of movement. So there's reason to believe that Joe Kelly has more potential than his numbers indicate.

However, delving into a more quantitative approach, despite his well below average 4.62 FIP in his 10 starts for Boston, Joe Kelly produces .8 RA-9 WAR. Extrapolated out to a full season and accounting for potential rounding error, by that metric, we can expect Kelly to be worth 2.2-2.6 RA-9 WAR over the course of a full season. So where does the difference between that and 1.1 FIP-WAR lie? There are two possible sources- Babip and LOB%. A quick glance at his stats shows that his LOB% was, if anything, below average, and his .237 Babip jumps out as unsustainable. However, Kelly generated plenty of free outs with his 13.3% pop-up percentage, and his career 52.4% ground ball rate allows for plenty of double pays to erase the baserunners he does allow. So while Babip regression is obviously due, the double play tendencies play down his already below league average career Babip of .290. Thus, Joe Kelly is the type of pitcher who is underrated by FIP-WAR. In Kelly's career 327 innings pitched, he has accrued 4.4 RA-9 WAR and only 1.9 FIP-WAR. The discrepancy is big enough and the sample size is large enough to suggest that simply looking at Kelly's FIP-WAR does not fully demonstrate his on-field contributions.

As has been pointed out by many, Boston has prioritized ground-ball pitchers to put in front of what will likely be an above average infield defense in order to maximize value. The same holds true for Joe Kelly, who had the 9th highest GB% of qualified starters last year. As he pitched in front of a stellar defense in St. Louis in 2012-13 that allowed him to greatly outperform his peripherals, he should have a similar quality defense behind him in Boston. Using his performance to date, Kelly's 4.4 RA-9 WAR in 327 innings extrapolates out to 2.4 RA-9 WAR over 180 innings this year, and given the freak nature of his injury last year, that projection doesn't seem overly aggressive. While I acknowledge that method is an imperfect projection, utilizing the same technique with FIP-WAR leads to a 1.0 Projection, just below Steamer. Basically, the key takeaway here is that beating FIP is a skill, one that, as shown throughout this post, Joe Kelly appears to have.

Thus, in acquiring a starter and moving Joe Kelly to the pen, the Sox would be replacing a 2+ win pitcher with a (presumably) 4+ win pitcher. While clearly that is not a small upgrade nonetheless, we would be paying the full 4 win pricefor a 2 win upgrade. Whether that price is a 9 figure contract or a slew of top prospects, I don't feel that is a necessary move.