Let’s Talk About Joe Kelly Fastballs

Joe Kelly has been a hot topic of discussion since he was acquired last year. It would be nice if he had an up and down season, but it’s simply been bad. Ian York talks about Joe Kelly’s fastballs, and how he can improve his pitching by improving his usage of them.

Watching Joe Kelly pitch, it’s hard to understand why he isn’t more effective than he has been this year. As well as a decent curve and a solid changeup, Kelly throws a set of very, very fast fastballs, often with dramatic movement. In fact, Kelly ranks third in the league this year (for starting pitchers with 80 or more innings) for fastball velocity, averaging 95-mph on his four-seam fastball, and fourth in the league for his two-seam, averaging 95.4-mph.

In spite of his overwhelming raw “stuff”, though, Kelly has not been a good pitcher at all. Damian Dydyn has recently summarized Kelly’s struggles, but the bottom line is that only his changeup and his two-seam fastball (or sinker) are remotely effective pitches. Fangraphs lists his pitch values as follows; we can also show the number of hits and balls per 100 pitches that each pitch type gives up (MLB averages for hits and balls/100 are shown in brackets beside Kelly’s numbers):

Four-seam fastball
Two-seam fastball
Fangraphs pitch value
Hits/100 pitches
7.0 (5.8)
6.2 (7.1)
6.6 (5.6)
7.1 (4.8)
4.2 (6.3)
Balls/100 pitches
34.8 (33.5)
35.4 (34.2)
31.1 (31.1)
37.4 (34.8)
43.1 (32.7)

(For comparison, a pitcher in the top ten for fastball quality, like Max Scherzer, gives up about about 4.8 hits and 26 balls per 100 pitches.)

Like most pitchers (especially those who can throw 100-mph) Kelly mainly throws fastballs (about 70% of his pitches), but even when he throws a different mix, there is little obvious correlation with success. Here are the outcomes for each of his games this year, compared to the pitch combinations he used:

While he has had some very good games in there, it doesn’t seem that there is any particular pattern of pitches that leads to good outcomes.

As a side note, as Damian pointed out in his recent article, the categorization of Kelly’s fastballs into “four-seam” and “two-seam” by PITCHf/x is arbitrary. For most pitchers, a four-seam is the faster and straighter pitch of the two, and there is a more or less clear distinction between the two clusters of pitches. For example, as we showed here, Rick Porcello’s fastballs this year show two clear groups, even though they are less well separated this year than in previous years. For Kelly, though, there is essentially no difference in velocity of the two pitch types. PITCHf/x seems to have arbitrarily decided that pitches with more than about 4.5 inches of break must be two-seams, regardless of speed:

At the major-league level, a pitcher with great velocity but little movement on his fastball is nearly certain to get hit hard, and so there has been a tendency to assume that Kelly must be throwing a straight fastball with no movement. That is not the case. Even Kelly’s fastest four-seam fastballs have considerable movement, easily in the same range as more effective pitchers, and the pitches that are called two-seams often have extreme movement (for a fastball):

Based on their speed and movement, Kelly’s fastballs should be dominating pitches, yet they are actually considerably worse than average. One reason is their location. Here, we are looking at the strike zones for left- and right-handed batters from the umpire’s viewpoint. The strike zone, as umpires called it in 2014, is the grey polygon, and the distribution of Kelly’s fastballs are shown as the shaded contour lines:

Unlike most pitchers, whose fastballs are generally clustered in one or two specific locations near the edge or bottom of the strike zones (for example, see Rick Porcello’s and Wade Miley’s location charts), Kelly’s fastballs are centered on the middle of the strike zones, and smeared broadly across them. Rather than the pitch itself being ineffective, Kelly does not seem to be able to precisely locate his fastballs (perhaps partly because of the amount of movement they show), leading to too many pitches in the batters’ wheelhouse.

When Carl Willis joined the Sox as pitching coach partway through the season, one of the goals he suggested was to have Kelly slow his pitches down, in the hope of improving their location. Is this a reasonable approach for Kelly? To get a sense of how effective his slower and faster fastballs are, we sorted his four- or two-seam fastballs by speed, and used a running average over groups of 200 pitches to see how many hits per each speed group yielded:

Clearly, so far this year his fastest pitches have been the hardest to hit. In the case of the two-seam fastballs (which, remember, are defined by PITCHf/x as being those fastballs with the largest break length), there are also a set of slower pitches that are moderately effective, but for both pitch types, those that are just about his average velocity (around 95.5-mph) are hit the most often, while the fastest pitches (those at 96-mph or more) seem to be much harder to hit.

Was Willis completely backward, then? Not necessarily, because here is the other half of the story:

Those very fast pitches may be hard to hit, but they are also apparently hard to locate, because they are much more likely to be balls than a slower pitch.

Kelly this year seems to be be trapped by his fastballs. If he throws them as hard as he is capable of, they are less likely to turn into hits, but more likely to become walks. If he slows his fastball down, improving his ability to throw strikes, the pitches become much more hittable.

For his two-seam fastball, it is possible that there is a sweet spot at the lower end of the speed range; below about 94.5-mph, his two-seam fastball is moderately hard to hit, and not too likely to be a ball. There does not seem to be a similar sweet spot for his four-seam at the moment. The good news may be that it is not far away; if Kelly can either improve his four-seam’s location, or reduce its hittability, just a little, then there might be a sweet spot around 96-mph where his four-seam could achieve the dominance it seems that it should have.

The other important point is that pitches should not be looked at in isolation. As with Porcello, if Kelly can find more separation of his two- and four-seam fastballs in velocity and movement, or if his changeup or breaking pitches were more common and reliable, all his pitches might become better through improved deception.

Pitchers like Joe Kelly, who can throw 99-mph through the course of an entire game, and who have that amount of movement on their pitches, can practically be counted on the fingers of one foot. Perhaps Kelly will never learn to completely harness his stuff. But he is already very close to doing so, and if he does, he has the potential to be a very good starter. If Kelly can reach his potential, a few more bad games in this already-written-off season would be a small price to pay.

Ian York has written about rookie struggles, an impressive start by Eduardo Rodriguez, Mike Napoli and the effect the strike zone is having on him, and the effect of better bullpens on offensive strategy.

Follow Ian on Twitter @iayork.

Check out Brandon Magee‘s weekly minor league report for the week of 8/7 and Ian’s look at Henry Owens’s debut.