The 2015 season was a disappointment for the Boston Red Sox, and that disappointment started with the rotation. The team made a step to remedy that this offseason when they signed David Price to a long-term contract. However, they will still send out four-fifths of 2015’s rotation, and Ian York looks at the PITCHf/x data to try answer the question: What can the Red Sox can expect from Joe Kelly in 2016?
We keep asking that question, and it’s probably time to admit that we just have no idea. Kelly has tremendous pure stuff – fastballs up to 100-mph, great movement – but his mediocre results have rarely reflected his raw talent. At 27-years-old and with four major-league seasons under his belt, youth and inexperience are not excuses any more.
And yet, Kelly finished the 2015 season with an impressive run of games in which he earned 8 of his 10 wins. Hope springs eternal: Is it possible that these games represented a step forward for Kelly, or are they more likely to be pure luck, like other decent stretches he has put together in the past?
I split Kelly’s season into two parts: His first 16 games through the end of July, during which he had a 2-6 record and a 5.94 ERA, and the nine games he pitched from August on, when he went 8-1 with a 3.00 ERA. For each of these periods, I looked at the effectiveness of each of his pitch types, looking at total bases per 100 pitches, and balls per 100 pitches.
On the total bases graph, the slider markers for first and second halves overlap each other perfectly, hiding the first half marker. His slider, change, and curve had fewer balls/100 pitches in his post-July stretch – his curve dramatically so. However, the curve is his least frequently-used pitch, and he only threw 70 of them in that period, so this is at least partly a small-sample-size anomaly.
More interestingly, Kelly’s fastballs were quite a bit more successful in the post-July period. In the first part of the season, he gave up 10.2 TB/100 on his fastballs; in August and September, this dropped to 6.8. Although the number of balls on the pitch increased slightly (from 35.8 to 39.9 balls/100 pitches), his fastballs went from being a liability to a significant strength. For comparison, his second-half fastballs were about equal in value to Zack Greinke’s (7.3 TB/100, 34.8 balls/100), and better than Jon Lester’s (8.2, 37.1).
It is worth noting that Kelly’s “fastballs” include both four-seam and two-seam versions of the pitch. Although he probably does throw varieties with different grips, they blend seamlessly into each other and it is virtually impossible to find a reasonable separation between them; for these charts, I have pooled the two together.
Kelly’s fastballs seem to have been much more effective in the second part of the 2015 season. Is there any reason to believe that this reflects a real difference? We can look at his pitch usage over the season, as well as looking at the speed of his pitches in individual games, to see if anything changed. The vertical line shows the one-month gap when he was sent to the minors.
In the charts above, four-seam and two-seam fastballs are both colored red, to indicate that they should probably be considered as a single broad unit. The upper chart shows pitch usage per game. Note that in his final ten games, Kelly reduced his fastball percentage in favor of his changeup and, especially, his slider. (It is best to ignore the final game; Kelly pitched only 2.1 innings before being pulled with a sore arm that turned out to be inflammation of the shoulder when checked by MRI.)
The lower chart shows pitch speed and clusters. It is easy to see that the slider/changeup cluster is larger in the last set of games (again, ignore the partial last game).
There are several other interesting and potentially important points from the speed chart. First, Kelly slowed his fastballs down by about 0.5-mph on average, starting around the time he was recalled from the minors on July 23: 96.2-mph average before being sent to Pawtucket on June 23, 95.7 mph after his return. Second, he began to vary the speed of his pitches, especially his fastballs and changeups, around the same time. This is easiest to see on his September 4 and September 9 games, where his fastball clusters cover a range from the low 90s to close to 100-mph. Earlier in the year, most of his fastballs clumped tightly over a small range of speeds in the mid- to upper-90s. For a more objective measure, the standard deviation of his fastball velocity went from 1.42 before being sent down, to 1.69 afterward.
There’s evidence, then, that Kelly had success when he increased the number of his secondary pitches, slowed his fastballs slightly, and varied the speed of his pitches. Is this chance, or is there any reason to believe he may have done this deliberately?
In fact, all three of these changes were specifically targeted by Kelly’s pitching coaches. One of the first things Carl Willis noted about Kelly was that he could benefit from slowing his fastballs down a little. Willis also changed Kelly’s pitch selection, according to Peter Gammons. And Red Sox pitching consultant Pedro Martinez gave Kelly a pitching lesson that emphasized changing speeds on his pitches.
So Kelly’s pitching coaches recommended a set of changes for Kelly; Kelly made these changes; and his effectiveness improved dramatically. That gives room for optimism, for anyone who hasn’t been burnt by unwarranted optimism over Kelly in the past. At this point, we will take a wait-and-see attitude to Kelly in 2016.