The Same and Improved Joe Kelly

As of March 16th of this year, everything was decidedly not sweetness and light for Boston right-hander Joe Kelly. On that day, he recorded his third (and ultimately final) start of spring training, over the course of which he conceded three runs in just 2.2 innings, bringing his spring ERA to 11.05. Worse yet, he was compelled to leave that start due to biceps tightness in his biceps (i.e. the place where that kind of tightness is most commonly found). The outlook was sufficiently grim that managing editor Dave Cameron was forced to publish a post here considering other starting possibilities for the Boston Red Sox.

Following a retroactively dated trip to the disabled list and a pair of minor-league spring-training appearances, Kelly made his season debut on Saturday. It’s hard to know what Kelly’s expectations were or what the organization’s were, but “low-ish” is an objectively reasonable assumption. If nothing else, there had to be concerns regarding Kelly’s endurance. Of the two appearances he’d made since leaving his spring start with an injury, his highest pitch count was 78. “Ideally he’d have another outing to build arm strength before an MLB game,” John Farrell said in the presence of Providence Journal reporter Tim Britton. That ideal scenario did not become a reality. Instead, Kelly’s next appearance was Saturday’s.

There were reasons, in other words, to expect the worst for Joe Kelly’s start on Saturday at Yankees Stadium. In reality, however, Kelly’s results from that start were actually the best. Not the best in every sense of the word, but certainly among the best so far as Kelly’s major-league career is concerned. He allowed just one run over 7.0 innings. He posted the lowest single-game FIP (41 FIP-) of all his starts ever. And another thing he did was to surpass his previous single-game strikeout mark. Previous to Saturday, he’d recorded six strikeouts in a single game on seven different occasions. On Saturday, however, he produced eight strikeouts (i.e. two more than ever before).

At some level, it’s not surprising to find that a pitcher of Kelly’s talents produced such an excellent game. Among the 185 pitchers last year to throw 50-plus innings in a starting capacity, Kelly recorded the 12th-highest average fastball velocity. Generally speaking, arm speed correlates pretty strongly with swinging strikes and (by virtue of that) strikeouts. By way of illustration, consider: the 11 starters with a higher average velocity than Kelly in 2014 produced a 23.2% strikeout rate collectively. Kelly? Just a 15.9% mark — or the 148th-best such figure among those same 185 pitchers. Jeff Sullivan considered this odd relationship between Kelly’s plus arm speed and distinctly not-plus ability to produce swings and misses back in 2013. So when that was also the case in 2014, it wasn’t a surprise.

What was different, then — if anything — about this version of Kelly from Saturday? Let’s begin and also probably end with an examination of his repertoire. Below is a table featuring some data regarding Kelly. Below that is an explanation of the data.

The first column denotes the individual pitches in Kelly’s repertoire, which remain unchanged since last year. He hasn’t added or subtracted any pitches entirely, in other words. It’s the same collection of pitches. The group of three columns after that features Kelly’s pitch data — usage (Usg), velocity (MPH), and swinging-strike rate (SwStk) — from 2014. The next group of three columns features all that same data from Kelly’s start on Saturday (Sat).

You’ll immediately notice that certain data points are highlighted here. Let’s begin with the yellow highlights. The highlighted figures on the left (95.8 and 95.4) represent the average velocities of Kelly’s four- and two-seam fastball, respectively, from 2014. The figures on the right (96.5 and 96.0) represent Kelly’s average velocities for those pitches on Saturday. The figures on the right are higher which means either (a) Joe Kelly is throwing harder now or (b) the PITCHf/x cameras at Yankee stadium were miscalibrated slightly — a thing that can happen.

Checking on point (b) is a bit tricky. Kelly faced Adam Warren on Saturday, who actually threw his fastball roughly 2.5 mph slower over the course of his start than he had in 2014. That might reveal something about the Yankee Stadium PITCHf/x equipment except for the fact that Warren threw exclusively in relief last year, in which capacity we’d expect him to throw harder. It’s possible that comparing a reliever from this game to that same reliever from 2014 would add something conclusive. The point is laborious and tedious, however — and also maybe unnecessary.

Unnecessary, maybe, because of this: now direct your attention to the reddish highlights. The left reddish highlights denote Kelly’s slider velocity and swinging-strike rates from last year; the right reddish ones, that same data from Kelly’s Saturday start. Here’s what one finds: that, according to PITCHf/x, Kelly threw his slider with a bit less velocity and also recorded swinging strikes at more than twice than rate that he did last year. (Eight of the 18 sliders Kelly threw resulted in whiffs.)

Not because it’s necessary, but because it’s pleasant, here’s Kelly whiffing Chase Headley in the fourth inning:

And then doing it again later in the at-bat:

And that throwing that second pitching again in slow-motion:

Kelly threw those two sliders at 85-86 mph. Overall, his slider sat at 86.0 mph on Saturday. He threw his average slider in 2014 at 87.0 mph. What are we to conclude from this? One of two things: either that (a) Joe Kelly is throwing his slider slower now or (b) the PITCHf/x cameras at Yankee stadium are miscalibrated slightly. The thing is, we already considered this possible miscalibration above — but in that case, the cameras were hypothetically adding to Kelly’s fastball. It’s unlikely that whatever error is providing faster reports on Kelly’s fastball is also providing slower ones on his slider. Kelly’s curve also produced a lower sitting velocity — suggesting that problem isn’t a conflation between Kelly’s two breaking pitches.

Were Occam here and were he to apply his razor to the case of Joe Kelly, he’d settle upon two points. First, even allowing for a miscalibration, he’d conclude that there’s probably a greater velocity separation between Kelly’s fastball and breaking pitches than in the past. Second, he’d observe that, regardless of whatever else is happening, Kelly induced more whiffs on his slider than previously. They’re all pitches that Kelly has had before, so it’s nothing new. The speed at which they’re being thrown is just maybe a bit different — and the results, definitely better.

In conclusion — and also for no reason — here’s a rich and compelling example of Joe Kelly’s two-seamer, thrown at Brett Gardner’s front hip in the sixth inning for a strikeout:

And also in slow-motion, again:

Data courtesy Brooks Baseball was very helpful for the composition of this post.