Changing Speeds: A Pitch by Pitch Breakdown

The 2021 MLB playoffs have, at this late stage, seemingly been dominated by pitching considerations. Load management, relief pitcher usage, and cherry-picked matchups are all at the forefront of managerial considerations and decisions. Pitcher performance after all has a lot to do with those matchups and usage, particularly after a long 2021 season on the heels of a shortened 2020 campaign.

It seems to me that one of the oft-cited indicators of performance mentioned by announcers these playoffs has been how well a pitcher changes speeds. Should a pitcher be moving through their opponent’s lineup efficiently, you can seemingly assume that one of those individuals covering the broadcast will laud that pitcher’s ability to mix their pitches, change a batter’s eye level, or change speeds – all familiar turns of phrase for postseason viewers.

The discussion of changing speeds does beg the question: how much do pitchers in actuality change the velocity of their offerings from pitch to pitch? Does a relatively high or low amount of velocity variation correspond to stronger performances? The playoffs don’t offer an ideal context for pondering these questions, given short outings, small samples of starts, and selection biases, but data from the regular season might shed some light on the value of that “changing speeds” phrase.

In search of some clarity on this topic, as well as more data, pitching performances for 2021’s innings pitched leaders were drawn upon. The group of pitchers was limited to the top 10 given that that small group alone threw 30,682 pitches in 2021, or roughly 3,000 apiece. That ~30,000 row figure is just about what Baseball Savant will allow to be downloaded in one go, so I stuck with the top 10. Just four pitchers threw 200+ innings in 2021, but all of the top 10 in innings pitched threw at least 185.

Very simply, I took the absolute average pitch to pitch velocity change over the course of each player’s ~3,000 total pitches in the 2021 regular season. For example, should a pitcher throw just three pitches with velocities of 90, 80, 85 then their absolute average speed change across those three offerings would simply be 10 (90 to 80) and 5 (80 to 85), or 7.5 mph, on average.

As another example, a pitcher that throws pitches with velocities of 90, 90, 90, 90, 75, 75 (0, 0, 0, 15, 0 in absolute pitch to pitch velocity changes) in that sequence averages just 3 mph on average in velo changes from pitch to pitch. Meanwhile, another pitcher that uses the sequence 90, 75, 90, 75, 90, 90 (15, 15, 15, 15, 0) threw the same velocities in the 6-pitch sample but had an absolute average velocity change of 12 mph.

This isn’t a perfect method, of course, as there is plenty of nuance neglected (not to mention, variation in velocities between a final pitch in one outing and the first pitch in a subsequent outing are meaningless), but it achieves a very simple view of how much velocity changes for a pitch from offering to offering.

Below is a table of those innings pitched leaders in 2021 and some corresponding speed change metrics.

Among even this small group of pitchers, there is a wide range of velocities for pitches thrown. Adam Wainwright‘s arsenal in 2021 included pitches with as much as a 30 mph difference in velocity. Frankie Montas, meanwhile, had a particularly consolidated mix of pitches, using velocity as a measure of disparity. Across these ten pitchers, ~20mph in maximum and minimum pitch velocities appears typical.

From the perspective of absolute average velocity change, Wainwright leads the way. On average, the “next pitch” he threw at any given point in 2021 was 8 mph faster, or slower, than what he had just thrown. This ten pitcher sample is tiny, and suffers from obvious selection issues, but that seems to be a considerable number nonetheless. On the other end of the spectrum, Sandy Alcantara‘s absolute average velocity change from pitch to pitch was the smallest among this group, checking in at under 4 mph.

A reasonable next question might be “does mixing pitch speeds relate at all to overall performance?” If one is to take postseason announcers’ words to heart, mixing speeds is clearly related to success on an outing to outing basis.

Obviously, it is inadvisable to compare pitchers to other pitchers by such a metric alone, as there are countless omitted variables in such a juxtaposition: one cannot say that changing speeds makes Adam Wainwright any better or worse than Sandy Alcantara while neglecting the many other proven factors that motivate the value of those individuals.

What might be a slightly better exercise is to examine the relationship between changing speeds and success for each pitcher’s individual starts as opposed to making a comparison across pitchers. In this instance, as a proxy for success, I have used the game score metric redeveloped by Tom Tango.Across these 10 pitchers’ ~300 starts, there was very clearly no relationship between game scores in individual starts and average pitch to pitch velocity changes. For individual pitchers though, there were some cases where changing speeds seems to relate somewhat to start performance. Below are the two most drastic examples illustrated in scatterplots.

For Zack Wheeler, mixing speeds didn’t seem to correspond with success in 2021. On the other hand, game scores for Frankie Montas were positively associated with greater absolute speed changes. It is worth reiterating though that for most pitchers in this 10 pitcher group, there was essentially zero relationship between game scores and in-game speed mixing.

Getting additionally granular (across these 10 pitchers at least), there was no relationship between individual pitch velocities relative to prior pitches and corresponding exit velocities for all pitches put into play, exit velocities here being a proxy for the relative success of a pitch. Below is a scatterplot which illustrates exit velocities for all batted balls allowed by those top 10 pitchers and the corresponding velocity change relative to the prior pitch.

The plot above suggests that how fast a pitch travels relative to the prior pitch thrown doesn’t correspond to exit velocities for pitches put in play.

Setting aside the value of changing speeds on a pitch to pitch basis, the following three histograms more generally depict the frequencies of various velocity changes. The first histogram covers pitches from all 10 pitchers covered here while the next two cover Alcantara and Wainwright, whose differing absolute average velocity changes (3.85 and 8.18, respectively) offer interesting variations in the histograms which represent their pitches.

To recap, changing speeds, evaluated in this way and for this tiny subset of pitchers at least, does not appear to significantly relate to performance. That said, there is some relation for some particular pitchers in 2021, namely Zack Wheeler of the Phillies and Sandy Alcantara of the Marlins, but in differing ways. On a pitch level, velocity relative to prior pitch velocity did not correspond to greater or lesser exit velocities. None of this supports or refutes the value of changing speeds in a particular game or, more importantly, in the playoffs, but it does offer some context for how pitches change speeds from pitch to pitch.