Editor’s Note: After an almost year-long hiatus, I am dusting off this site and once again sharing content. The plans to share more NECBL analysis from the summer of 2021 ended up getting lost in the midst of my senior year, Notre Dame Football, and a variety of other factors. However, armed with a renewed commitment, more projects to share, and a bit more freedom on the weekends, I hope to post some form of analysis on a semi-weekly basis. Now, onto pitch framing analytics!
This summer I once again had the privilege to run baseball analytics for the Vermont Mountaineers in the NECBL. Thanks to manager Mitch Holmes, I was able to be in the dugout as a bench coach and work directly with the players and coaches. The 2022 Mountaineers consisted of an extremely talented group of players (they matched the NECBL regular season wins record) and probably the best dugout I’ve ever been a part of. In terms of analytics, my role in the dugout allowed me to expand on my 2021 projects in a variety of ways, most of which I hope to cover in future posts. Overall, it was truly a wonderful summer of baseball.
Today I specifically wanted to highlight catcher framing, which was one of my favorite analyses from the summer. Given how unreliable NECBL umpires can be (more on that in a bit) analyzing our catchers’ ability to steal strikes gave us the opportunity to identify areas of improvement and leverage this ability in pitch calls and lineup decisions. I will segment this post into three main sections: 1) demonstrating the need for such analysis, 2) explaining the process of analysis, and 3) examining the results of the 2022 season, including highlighting use cases, areas of improvement, and next steps.
The Need for Catcher Framing Analysis in the NECBL
As mentioned above, the strike zone in the NECBL posed a particular challenge to batters but an enticing opportunity for pitchers and catchers. To demonstrate visually, take a look at the image below. Plotted in black is the ‘true’ strike zone’, with the actual plate dimensions inputted and an assumed vertical zone from 1.5 to 3.5 ft. This represents what a perfect strike-zone would look like. The zones in the blue/red outline show the realized strike zones from D1 baseball over the last five seasons (right zone) and every Mountaineers game this year. (left zone) These zones were created by applying a generalized additive model (GAM) over pitches not swung at in order to estimate the probability of a strike. The line between the red and blue areas represents the 50% line; pitches outside of that had a less than 50% chance of being called a strike while pitches inside that line had a greater than 50% chance of being called a strike based off actual ball/strike calls (this strategy is borrowed from a FanGraphs article on pitch framing which I highly recommend).
It is fairly apparent from these zones that the strike zone which collegiate players work with in both summer and school ball does not match the rulebook-defined zone. Especially in the NECBL, pitches that were almost 6 inches off the plate were being called strikes at about a 50% clip. Furthermore, notice the degree of variance in the 75% area (the red) and 25% area (the blue). Not only were the zones expanded, the consistency in expansion varied umpire to umpire.
This discovery offered a clear advantage for our pitching staff and catchers, but also emphasized the importance of having a catcher who could routinely frame pitches. Despite the expanded zone, a catcher who did not stick pitches or frame his body correctly would still be at a disadvantage, while a particular skilled strike-stealer could see additional success with the realized NECBL zone because of how varied the strike calling accuracy was. It seemed the value of pitch framing in the NECBL was larger than Division I baseball and MLB due to these expansive and inconsistent zones. However, the ability to measure individual catcher performance was needed in order to realize this additional value.
Side note: One of the main difficulties in this analysis is untangling actual umpire skill from the effect of pitch framing. I have a hypothesis, based only on visual evidence, that NECBL umpires were particularly influenced by outside factors (catchers, coaches, etc). Furthermore, the Mountaineers catching staff, empirically, was much more successful at stealing strikes than our opponents’ catching staffs, which I will demonstrate in the third section. This does mean, however, that the Mountaineer zone presented above may be biased towards a larger strike zone based on the sample being skewed by our pitch framing. To alleviate this concern, I utilized the Division I zone presented above, which was built off of five years of D1 Trackman data (>1M pitches), for all performance analysis. This hopefully eliminates the bias of pitch framing, as it includes equal representation from poor framers and elite framers across the D1 level over a five year span.
Measuring Individual Pitch Framing
In order to quantify individual performance, I decided to measure both the total strikes added of a catcher as well as the run expectancy those strikes mapped to. After I filtered out pitches that were swung at, I then ran a few loops through the dataset. The first loop labeled pitches as either a ‘true’ strike or a ‘true’ ball, depending on if they were located in the rulebook strike zone or not. I also ran a similar loop through the data set labeling pitches as either a Division I strike or Division I ball depending on if they were within the 50% line of the realized D1 zone from above.
With every pitch labeled in this fashion, I then compared the actual call with the ‘true’ value. If a pitch was called a strike, but was a true ball, I labeled that pitch as a Strike Added. If a pitch was the opposite, a true strike called a ball, I labeled that pitch as a Strike Lost. I did this for both the ‘true’ strike zone and the Division I strike zone.
After identifying Strikes Added and Strikes Lost, the next step was to map run expectancy values to them. To do so, I utilized a run value matrix from the Hardball Times they created in 2014 from this post (shown below). With this, I could properly differentiate between pitches that were framed in low impact situations versus those framed in high impact situations.
Now, I admit the matrix is old, from MLB data, and supposed to be dynamic. One of the goals I have is to recreate this matrix for the NECBL and rerun the catching analysis from the summer. Furthermore, the article linked above also discusses the implications of hitter quality, another dimension which could be brought in to refine the run valuation. However, this matrix provided a rough estimation of the run value of pitch framing and for the purposes of in-season use, that was good enough. The run values were mapped to the corresponding counts with Strikes Added or Strikes Lost, applying either a positive or negative run value, respectively. With added run values to the data set, it’s time to actually evaluate the Mountaineers catchers.
Evaluating Pitch Framing: Examples, Issues, Next Steps
In this section, I want to highlight the pitch framing of Christian Pregent, the catcher at Stetson University. Christian spent the summer with the Mountaineers and, in my opinion, was the best defensive catcher in the NECBL. Using him as an example, I will walk through the application of this pitch framing metrics.
Evaluating pitch framing provides both player development and strategic advantages. From a player development perspective, it can provide an intuitive analysis of where an individual catchers excels or struggles in framing pitches through visualizations. These visualizations, paired with game film, provide opportunities for player development and individualized framing drills.
Let’s examine this play development lens a bit further by taking a look at Christian Pregent’s data visually. An individualized catcher zone, similar to the league umpire zones above, can be generated by running a GAM function over all pitches received by a catcher. Below is Pregent’s zone over approximately 1800 pitches (~170 IP) compared to the Division I zone:
Pregent’s zone clearly is a bit more expansive than the Division I zone, with the 50% threshold occurring approximately 9 inches off the edge of the plate in both horizontal directions. While the 50% GAM visualization provides a quick comparison, the function can also be used to provide a more granular analysis in the form of a heat map, as demonstrated below.
This heat map also includes a super imposed heart and shadow region of the strike zone. With a more granular view, the true nature of Pregent’s zone becomes apparent. This chart again emphasizes the ability to add strikes on the inside and outside portions of the plate, but it also shows that the upper and lower regions of the zone have sharper cutoffs than the outer regions. Now, this visualization combines both left-handed and right-handed batters, making it difficult to determine if Pregent excels on both the inner and outer portions of the plate, or if it is just one direction (ie outside to both right-handed and left-handed hitters). Additionally, it includes both fastballs and offspeed pitches, which often are framed using different techniques. The graphics below hone in on these different dimensions, offering a unique zone for filtered pitches.
Examining the LHH v RHH difference, it becomes clear that where Pregent really excels is on the outside corner of the plate. While the inside corner also extends beyond the real zone, the outside edge of the zone is around 2-3 baseballs more than the actual zone, a huge advantage for any pitcher that Pregent catches. Conversely, the offspeed v fastball comparison demonstrates more of a high-low differentiation. The offspeed zone exists approximately 1-2 inches higher than the fastball zone, demonstrating a proficiency at framing high offspeed pitches, but not as much the low ones. Examined from a player development standpoint, the main focus on improvement would probably be those low offspeed pitches, especially given the center-cut peak at the bottom of the offspeed zone. All in all, Pregent’s framing zones extend well beyond the real strike zone and demonstrate top-end receiving skills.
Ultimately, how does Pregent generate those expanded zones? What techniques/mechanics does he employ to gain such an advantage? The question of how intrigued me over the summer and allowed me to pick the minds of players and coaches who were far more technically knowledgeable about player mechanics, technique, and development than myself. I had many discussions with both Pregent and Mountaineers catching (and hitting) coach Ural Forbes over the course of the summer trying to isolate the cause of Pregent’s success. Ultimately, most of the conversations centered on body positioning for in-out framing and glove depth for high framing of offspeed pitches.
These techniques can best be explained using game film. Take a look at the pictures below of Pregent before and after a pitch. The call was a fastball outside. Notice how Pregent centers his body on the glove, which is on the outside corner, but angles himself slightly inwards towards the plate. This allows him to catch the pitch, which was about 7 inches off the plate, between his shoulders, with minimal glove movement and the glove extended back towards the plate. Because he positioned his body in this manner, the pitch is presented as hitting the corner, as opposed to being outside. If he had set up his body more centered with home plate, this pitch would have been caught outside his left shoulder, resulting in more glove movement and potentially a ball call.
Here’s another example demonstrating the same technique, but this time on a slider. Once again, Pregent sets himself up off the plate, but angled inwards, allowing himself to catch the pitch between his shoulders and present it more inward than it actually was. The pre-pitch body positioning sets him up for a cleaner frame. Since this is a slider, he pre-sets his glove lower, allowing him to catch the ball with an upward motion, naturally framing the ball in the zone.
Here’s two more examples of the same concept, this time inside and outside against a right handed hitter. As with the past two examples, the pitch is caught in the center of his body, eliminating unnecessary movement and presenting the pitch as a success.
In terms of high offspeed pitches, the technique was best explained to me as allowing the pitch to travel deeper towards the catcher’s body before catching it, thus allowing the ball to break more before the frame. This one is a bit tougher to see from the center field angle, but the sequence below demonstrates the difference in where the ball crosses the plate and where it is presented. Ideally, the difference between the two comes from the ball being caught deeper in the catcher’s chest rather than from dragging the glove downwards.
With that understanding of the technique behind pitch framing, let’s end by considering the strategic perspective of this information, more specifically, the ability to provide an estimation on the value of pitch framing, in runs. As mentioned above, for each strike added or lost, the associated run value based on count was applied to that pitch. Ultimately, these values can be aggregated and plotted to examine the total runs added (or lost) by pitch framing over the course of a season. Below is this analysis for Pregent’s 2022 season with the Mountaineers, with the graph on the left representing the real strike zone and the one on the right representing the Division I strike zone.
Even though some strikes are lost vertically against both zones, the evidence overwhelmingly indicates that Pregent’s framing nets positive strikes and runs for his team. Furthermore, the Strikes Added versus the Division I zone can be used as a proxy for Strikes Above Average, as the Division I zone is built from the aggregation of all Division I strike calls over the last five years. Therefore, with around 170 IP as catcher this year, Pregent netted the Mountaineers 120 Strikes Above Average, equivalent to about 11.9 runs saved just from pitch framing. As a rate, that would be 6.4 strikes and .63 runs added per 9 innings from pitch framing alone, a pretty remarkable pace. Even if hitting at a replacement level (hypothetically), the value generated from pitch framing alone would be enough to start Pregent behind the dish, especially if leveraged with pitching matchup data.
This aspect of the analysis provides more context to strategic decisions regarding lineups, bullpen usage, pitching matchups, and umpire relationships helps remove bias about catchers’ defensive prowess, a notoriously difficult skill to judge visually. Furthermore, there are plenty of opportunities for next steps. Ideally, taking individual batter, pitcher, and umpire data into account would make the strikes added and runs saved statistics slightly more precise. For instance, creating a GAM zone for each home plate umpire and then charting strikes added/lost against that zone would help separate pitch framing from umpire performance. Additionally, incorporating batter strength, pitcher strength, and base-out situations into the run values would help gauge value further. Finally, having access to glove positioning data (not tracked in the NECBL) could allow some analysis on pitch location as it crosses the plate and pitch location when caught. The difference between those two numbers could prove interesting in analyzing the technique of pitch framing a bit further.
To wrap up, pitch framing analysis was some of my favorite work to do over the summer and I was extremely lucky to work with such talented framers. Feel free to send questions, comments, etc through the ‘Contact Me’ page or by reaching out to me on Twitter (@j_messy2). And be on the lookout for more consistent content on here!
PS: I know it was a long post, but for those interested, here are the Mountaineers pitch framing numbers as a team.
Jacob Messineo's Analytics Blog 
One thought on “Revisiting the NECBL: Measuring Catcher Framing”