Over the last couple of weeks, my efforts have been spent breaking down the various aspects of swingmen, those pitchers with plenty of time as both a starter and reliever in the same season. The first entry focused on their aggregate data in both settings, revealing that as relievers the group-defined as having at least 10 starts and 10 relief appearances in the same year-improved in both ERA and FRA by approximately 0.7 runs. While the strikeout rate of these pitchers increased in the bullpen, the frequency of free passes issued stayed the same. Last week, we took a look at the dozen players since 1974 who fell into the bin of power starting pitchers but managed to undergo a complete 180 in the bullpen, becoming finesse pitchers; such an occurrence was not only rare in theory, but in practice as well. Here, I’ll be turning to looking at swingmen through a PITCHf/x lens.

What do we want to glean from the data? Prior research with swingmen was devoted more to outcomes, while PITCHf/x lends itself more to the inputs, the idea of results versus a process. Of interest is the velocity and movement of pitches in both areas, but the pitch breakdowns could also prove interesting. Perhaps, as expected, swingmen throw harder and with more movement in the bullpen, but might they also change up their approach, discarding a certain pitch and/or throwing one much more often?

First, our criteria must be clearly defined. PITCHf/x data from 2008-09 will be used, and the playing time parameters require a slight modification to account for the substantially smaller sample. For the purposes of this piece, a swingman will be defined as any pitcher who has a combined five starts and five relief outings since 2008, a total, not an annual figure. Unfortunately, this pitch-tracking dataset does not record fields similarly to Retrosheet in terms of denoting whether or not he who threw the pitch was the game’s starter, so an extra step must be taken to make this clarification. I chose to sync up the PITCHf/x data with that of the Retrosheet game logs, through the mapping of player identifications; the process essentially outputs the entire dataset for each pitcher in each game he started. From there, it became the much simpler matter of finding how many hurlers fit the aforementioned playing time criteria, and parsing out only the data for those pitchers.

With the procedure well defined, over 100 pitchers were returned, comprising a smaller sample than desired but still usable given that only two years of reliable data are available. Below are the velocities, movements and percentages for the four major pitches when the swingmen are both starters and relievers. Note that the percentages are relative to the four pitches below, and not every pitch thrown throughout the season was used:

As Starting Pitchers
Pitch         %      Velo    Horiz.  Vert.
Fastball     67.6    90.3     1.6     8.4
Slider       14.4    83.8     1.4     2.6
Curveball     7.4    76.2     2.4     5.2
Changeup     10.6    81.8     1.5     5.6

As Relief Pitchers
Pitch         %      Velo    Horiz.  Vert.
Fastball     61.2    91.0     3.4     8.2
Slider       17.9    83.5     1.7     3.0
Curveball    10.2    76.7     3.3     5.4
Changeup     10.6    82.1     3.2     5.7

What initially stands out is the fastball data: while velocity certainly increases in the bullpen, it is tough for me to imagine that I was the only one expecting to see a 2-2.5 mph uptick. Instead, at least as this sample suggests, pitchers “only” gain 0.7 miles per hour on their heaters as relievers. What they did not gain in velocity, however, they more than made up for in movement, barely missing a beat on the vertical front while more than doubling the horizontal component. In both of these situations, it is not a matter of two-seamers being lumped in and skewing the data, as I opted to solely include four-seamers.

What does this mean? Recall that movement in PITCHf/x terms refers to the way the pitch moved relative to a pitch thrown at the same velocity but with no spin. When the fastball is output at over eight inches vertically, it does not mean the pitch is rising, but rather that the same pitch albeit with no spin would fall that much more-as in the forces acting on the pitch keep it elevated more than if it were thrown without spin. So, put together, the fastballs were moving a heck of a lot more, and with increased velocity, even though not everyone managed to fall in line with Miguel Batista, who averaged around 90.5 mph as a starter and over 92 mph in the bullpen.

Of the four major pitches, only sliders lost velocity in the bullpen, while all experienced substantial upticks in horizontal movement. In note form:

  • All pitches were thrown with greater velocities, except for sliders;

  • All pitches were thrown with greater vertical movement, except for fastballs;

  • All pitches, bar none, were thrown with much more horizontal movement.

Given the small sample of the data in which to study, it is tough to conduct studies gauging the impact of velocity in conjunction to movement and vice-versa, to potentially determine if intuition’s suggestion that fantastic movement can increase perceived velocity of a pitch. Overall, the results above make sense, as pitchers should be able to throw harder in the bullpen, just as they should be able to throw their sliders and curveballs with greater snap and supination, adding movement. In the bullpen, these pitchers do not need to focus on any type of pace and can literally let loose, utilizing an “every pitch is my last” strategy. While I cannot recall studies correlating success to velocity or movement, it is certainly a valid proposition that success is more easily attained with electric stuff.

What does not make much sense in the data, however, is the pitch selection breakdown, as conventional wisdom dictates that most relievers end up in that role as a means of hiding their liabilities as starters, such as only boasting one or two above average pitches; in the pen, a less varied approach can work, and three batters can be effectively retired on a two-pitch repertoire, whereas the rotation requires a better mix of pitches and normally necessitates the usage of more than two offerings. It would then stand to reason that relievers rely more heavily on the heater with equally decreased rates of off-speed pitches, an intuitive idea that is contradicted by the data, which indicate that four-seamers are thrown more often in the rotation, while curveballs and sliders get used a little more often in a relief role.

One thought was that perhaps the group data is misleading in this case, since certain pitchers may throw four pitches as starters but only two or three as relievers; as in, one might go from all four pitches to just using fastball/slider/change or fastball/slider/curve. Further research suggested this was not really the case-in that the idea of it being misleading was overstated-since so few pitchers actually did scrap a pitch in the bullpen. Just 10 of the pitchers delivered a pitch more than 7.5 percent of the time as starters but under four percent as relievers. An example would be the changeup of Brett Myers, which he threw 10.1 percent of the time when in rotation but a mere 2.5 percent coming out of the bullpen. Another more extreme example was the changeup of Jose Contreras, thrown 13 percent of the time when he was pitching as a starter, and just 3.8 percent when relieving.

This sort of situation occurs, but not very frequently, and certainly not enough to dispel the data. An interesting comparison then would be the pitch breakdowns for swingmen relievers and the breakdowns for all relievers:

Pitch     Swing %    Relief %
Fastball    61.2       63.8
Slider      17.9       18.6
Curve       10.2        8.5
Changeup    10.6        9.2

By comparison to the larger group of which swingmen are a subset, those that go both ways throw their fastball less frequently, and are more likely to throw their secondary offerings, a characteristic that makes some semblance of sense given their time as starters.

Fusing all of this information together, as well as the results-based data from a few weeks ago, swingmen certainly sport different numbers in each role, but the deltas are not as significant as we might think-at least based on the data here and the outcome data from 1974-2009-with run-prevention changes along the lines of 0.7 runs and fastball velocities 0.7 mph faster. By guesstimation, it made sense that the numbers could instead be closer to 1.4 runs and 1.5-2.5 miles per hour. Alas, as we saw last week when discussing the odd role changers, not all swingmen are created equally and the definition is quite broad, categorizing pitchers based on playing time as opposed to approach or skill levels. Further binning could reveal much more telling data, but we can only go at the pace of the PITCHf/x dataset right now, meaning this is something that should be revisited five years out in order to see if anything meaningful changes.