How Do Pitch Characteristics Relate to Value?

An evolving phenomenon in baseball is that of pitch design. In general terms, pitch design entails making adjustments a to players’ pitch(es) based on that pitches attributes. This could be achieved a plethora of ways, from locating intentionally to better glean value from a pitches characteristics (placing high-spin fastball up in the zone), adjusting usage to complement another pitch (adding a changeup that’s horizontal movement is opposite that of an existing slider), by making physical adjustments to achieve a more effective spin axis, or incorporating any number of other adjustments.

I am no expert in pitch design, or frankly even particularly well-versed. For a couple strong examples of the thoughts and considerations that go into pitch design though, Michael Augustine has published several pieces on FanGraphs that I found to be really interesting and insightful. This post, meanwhile, does not wade into pitch design so much as it offers visual demonstrations of the relationship between pitch characteristics and their summarized values. Those relationships can in some ways offer a hand in informing pitch adjustment decisions. After a brief description of the data, the following figures should hopefully more or less speak for themselves.

The data which support the figures below were initially disparate and thereafter have been joined. For pitch characteristics (velocity, spin rate, horizontal movement, & vertical movement) data originated from Baseball Savant. For pitch values, data comes via FanGraphs’ Pitch Type Linear Weights, or Pitch Values, which summarize and standardize the value of any given pitch. The Pitch Value stat appeared to be a good catchall for performance of a pitch; you can read more about those it here.

Data for 175 qualifying starters over the last five seasons (2015-2019) was then joined, based on player name and season, so that data for pitch characteristics, via Baseball Savant, matched corresponding data for pitch values, via FanGraphs. The goal being to evaluate the relationship between the aforementioned pitch characteristics and the value that each pitcher got from using that pitch. Four-seam fastballs, changeups, sliders, and curveballs were each taken into account.

To illustrate those relationships, the data has been broken down by pitcher handedness, pitch type, and presented in scatterplot matrices. There are eight in all: four different pitch types and two alternatives of pitcher handedness. Below is the first matrix, which summarizes data for fastballs thrown by right-handed pitchers. A brief interpretation follows.

From the upper left to lower righthand corners, histograms represent the distribution of fastball value, velocity, spin rate, horizontal movement, and vertical movement, respectively. The scatterplots in the lower left visualize how each metric relates, while the numbers in the upper right offer the correlation between each metric. For example, the correlation between fastball velocity and vertical break is 0.67. The data is clearly noisy, but there are weak positive relationships between the fastball characteristics currently sought after (more speed, spin, and less vertical break, or “ride”) and positive pitch valuations. This next matrix covers fastballs from left handed pitchers.

It appears as though left handed pitchers haven’t been throwing quite as hard as their right handed counterparts, and correspondingly haven’t generated as much spin either. In the case of left handed pitchers, we find relationships that vaguely mirror those for right handed pitchers, but often not so strong. Data is even noisier for left handed pitchers across the board, in part due to the substantially smaller cohort.

These next six matrices cover curveballs, changeups, and sliders in that order, but in each case broken down first by RHPs and second by LHPs.

Here it appears that curveball spin was most strongly associated with curveball value, although the relationship still wasn’t terribly robust. Interestingly, and unexpectedly though, it appears that curveball spin more strongly correlated to horizontal movement than vertical.

Reaching concrete conclusions from matrices like the above is probably inadvisable, given a number of factors. For one, these data came from two disparate sources and it is not clear how large a sample supported any of the metrics in question. A fastballs average horizontal break, or a curveballs “value”, to use two examples, might not be very helpful if those metrics came from just a handful of pitches as a sample, which might be the case. Additionally, it is generally unfair to evaluate any given pitch without taking into account its location, the situation, or the catcher (not to mention the person it’s being thrown toward). Still, these matrices do a good job summarizing a considerable amount of data in a concise manner, while illustrating some key relationships between pitch characteristics along the way.