Howdy folks! Our Sequence Similarity Ratio (SSR) project received such a great reception online and at the 2023 Saberseminar, that we felt inspired to take a deeper look at sequencing. If you are just now coming across the project or need a refresher, I encourage you to check out our original post here. When we initially blasted the post on twitter in June, @jaketroch retweeted it and mentioned that he felt that we should consider evaluating sequencing through the lens of the count. For anyone who may be unfamiliar, a count is the current number of balls and strikes that a batter has in a given plate appearance. After chewing on a few ideas, we decided that we agreed with Jake! In this post, we will go into greater detail on this topic and we will share updated Overall SSR rankings.
Evaluating pitch sequencing and cutting it by the count was a tricky concept to wrap our heads around. We felt that the proper approach was to consider pitches thrown in a sequence that took place as of and after a given count. For example, when evaluating pitch type sequencing for a 3-1 count (3 balls, 1 strike), we included the pitch type thrown next as well as any other subsequent pitches thrown. In the example below, only the last three pitches would be included for evaluating SSR and the sequence would only be compared against other sequences that contained pitches that took place as of and after a 3-1 count.
We initially began this analysis with this methodology, but quickly realized that many counts’ sequences were being cut short due to batted ball events, walks, or strikeouts and thus limiting the length of sequences that were being evaluated. This was especially prevalent for counts that were near four balls or three strikes. So, we only included count specific sequences for sequences that had at least two subsequent pitches thrown. In the example below, the first 0-2 sequence would not be included because the plate appearance concluded on the 0-2 pitch. The second example would be included because the sequence, that contained pitches that took place as of and after the 0-2 count, consisted of two pitches.
With the methodology understood, let’s dive into the results. We were primarily interested in seeing which pitchers saw their SSR become more or less similar, given the count, when compared to their Overall SSR. We only included pitchers who faced at least 100 batters through 8/15/23. We also decided to not split by batter handedness as it would create many more elements to this analysis (but could be a great idea for exploration in the future). As mentioned in our original post, we feel like the SSR metric is only useful with context. Knowing how a pitcher’s pitch type sequencing patterns change, given the count, could be a very valuable insight for a batter.
Kicking things off, here are the results for pitchers whose Count Specific SSR were less similar than their Overall SSR. We are only showing pitchers who had the largest differences when comparing the Count Specific SSR to their Overall SSR.
There are a few pitchers that populate more than one slot, like Aroldis Chapman, Josiah Gray, and Wade Miley. Their repeated representation makes sense given the count progressions that they are holding reign on. Instead of diving deeper into one of these players, we were intrigued by Dinelson Lamet, formerly of the Boston Red Sox, and his pitch type sequencing approach during and after 2-0 counts.
It’s kind of odd to see that a pitcher significantly alters their pitch type selection when down in a count 2-0. If anything, you’d think that they’d double down on what they throw most frequently. In Lamet’s case, he actually deviated quite a bit from his typical approach. In the graphic below, you can see how Lamet’s pitch mix differed for the 2-0 count and forward pitch type sequencing when compared to his overall pitch mix. Once in a 2-0 count, his slider usage dropped by about seven percentage points, his sinker usage increased by about five points, and his usage of the 4-seam fastball increased by about 4 points.
When looking at two pitch pairs in these sequences, the shift we saw above was corroborated. Simply put, he was moving away from throwing sliders and choosing to throw fastballs more often.
Naturally, the next question was “why is this happening?”. To explore this, we plotted each pitch type’s location when thrown in and after a 2-0 count and for any count. We thought that the location of where the pitches were thrown, as well as the movement profiles, could be playing a role in deciding when to throw them.
Starting off with the slider, it’s clear that Lamet prefers to throw it low and typically outside the zone, or just biting the edge. Knowing this, it is reasonable to conclude that if the desired location is normally out of the zone, then throwing it when down 2-0 and risking a third ball would not be wise. Layer in knowing that his slider’s vertical and horizontal movement is below league average, it would appear that Lamet is trying to avoid hanging a juicy slider in the zone in a great batter’s count. The same can be said about his usage of the changeup. While only thrown at a very low clip, the idea is the same. He tends to throw it low and out of the zone and the pitch’s vertical and horizontal movement is well below league average. We believe that Lamet moved away from using these two pitches, once in 2-0 counts, for these reasons.
Moving on to the sinker, recall that once in a 2-0 count, Lamet increased its usage up to 32% from 27%. Looking at the plots, it is evident that he regularly locates his sinker low, but in the zone. Lamet also increased the usage of his 4-seam fastball to 28% from 24%. Similar to the sinker, the 4-seam fastball was regularly thrown in the zone, but higher than the sinker. While the vertical and horizontal movement of these two pitches are not special, it is Lamet’s ability to locate them for strikes that made them ideal choices when behind in the count 2-0. Unfortunately it didn’t seem to help much as Lamet was DFA’ed by the Red Sox after one appearance after being acquired from the Colorado Rockies.
We also looked at pitchers whose Count Specific SSR became more similar when compared to their Overall SSR. This group was equally as interesting given that they leaned further into their preferred pitch types. The results once again had players holding multiple count slots, like Kyle Hendricks and Bryce Miller. Seeing that we focused on the 2-0 count for pitchers whose Count Specific SSR was less similar, we decided to do the same for this group.
Bryce Miller, of the Seattle Mariners, popped up on this list several times. He’s been having a fantastic rookie campaign for the Mariners, so he was especially interesting to focus on. Similar to Lamet, once Miller was in a 2-0 count, he further leaned into his 4-seam fastball and sinker. Granted, his overall 4-seam fastball usage is already quite high at 65%, but his usage increased to 79%! His sinker was the only pitch type that also saw an increase in usage and each other pitch saw slight to significant decreases in usage.
Looking at two pitch pairs in these sequences, the shift towards the 4-seamer and away from just about everything else was further evident. Pairs that featured the 4-seam fastball generally saw an increase or were flat, unless paired with the slider, sweeper, or changeup. Interestingly, the slider and 4-seam fastball pair saw a slight increase in usage despite the strong overall drop in slider usage.
The reasoning for Miller’s shift towards increasing the 4-seam fastball usage could certainly be similar to Lamet’s. The location of his 4-seam fastball and sinker show repeated placement in the zone. I’d be remiss to neglect sharing that Miller’s walk rate of 5% is in the top 9% of Major League Baseball. Not only can Miller place his fastballs in the zone when needed, he doesn’t necessarily need to throw it over the heart of the plate. His pinpoint control allows for him to place the ball in less desirable locations that are still likely to be called strikes. The plots below, that detail his 4-seam fastball and sinker locations, help illustrate this point (we also suggest scrolling back up to compare to Lamet’s plots). Additionally, the vertical movement of his 4-seam fastball is interesting given that it has much less drop than a typical MLB 4-seamer.
Looking at the count specific plots versus the overall plots for the remaining pitches tell an interesting story. It is clear that he can place these pitches for strikes, but he avoids throwing them on 2-0 counts. Though, when he does throw them, they tend to be out of the zone. We imagine that this could be due to trying to avoid hanging a breaking ball or off-speed pitch in a favorable count for a batter. It will be very interesting to follow Miller’s career and monitor his sequencing - especially if he begins to mix in more off-speed and breaking balls.
Wrapping the Count Specific SSR results discussion up, Lamet and Miller ended up approaching sequencing once in a 2-0 count from different directions, yet did so in a very similar manner. Lamet typically throws sliders but moved to throwing more 4-seam fastballs and sinkers. Miller already throws a lot of 4-seam fastballs but further leaned into his best pitch. Would we see similar trends across the league where players ditch off-speed and breaking pitches for fastballs when behind in a count? It certainly seems like a good hypothesis. This analysis offers a glimpse into this and it would be fascinating to conduct similar deep dives into other counts, but that will need to wait for another day. Either way, we feel that this project and analysis offers a good template for developing situational pitch sequencing scouting reports. Heck, the White Sox could have used it last week when Luis Castillo pumped 47 straight fastballs to them!
Here are the updated Overall Sequence Similarity Ratio results split by the number of distinct pitch types that pitchers throw. All results are through 8/15/23.
Thanks again for checking out our blog post. If you have any questions or comments, please be sure to reach out!