The Benefits of Changing a Hitter’s Eye Level

There is an old adage in baseball that changing a hitter’s eye level pitch-to-pitch will lead to better outcomes for the pitcher. This makes sense on its face: compared to varying pitch heights and forcing a hitter to alter his bat path, throwing two consecutive pitches at the same height should make it easier for a batter to square up the ball. In a New York Times piece by Tyler Kepner, Mike Mussina discussed the importance of varying locations pitch-to-pitch to mess with the hitter’s eye, offering the example of throwing fastballs down and then countering with a pitch up in the zone. Kepner noted that the hitter’s eye would then be trained on a pitch higher in strike zone, affording the pitcher the opportunity to throw a curveball down to induce a groundball, or net a swing-and-miss. David Price has expressed a similar sentiment: “That’s always a big emphasis [for] me, just making sure I’m hitting spots with that fastball – two-seam, four-seam, both sides of the plate, moving it in, up, down.”

In research on the effect of eye level change on college hitters’ performance against fastballs, Higuchi et al. found that quick eye movement as a pitch traverses towards home plate has negative consequences for the hitter. This research was included in Driveline Baseball’s examination of hitters’ gazes when standing at the plate. On these pages in 2015, Jonah Pemstein looked into whether a pitch thrown at a different height than the one that followed it affected how umpires called the pitch at hand. Permstein surmised that this was indeed the case, with umpires less likely to call a pitch a strike at any height if the previous pitch was thrown at a different vertical location.

As I said up top, this all makes intuitive sense. But does it hold up to further scrutiny? The research I cited by Higuchi et al. only included six collegiate hitters and only considered fastballs. While their work was extremely thorough, its scope didn’t consider the hitter population many of us are most interested in (major league hitters) and only included fastballs at a time when pitches are leaning on breaking balls and offspeed pitches more than ever. Pemstein’s research looked at umpires, not hitters; his conclusions give us some confidence that behavior changes when pitchers vary their pitch location, but doesn’t provide insight into the strategy’s ability to flummox batters. I decided to delve into the data myself and see if there was any merit to this fundamental aspect of pitching strategy.

Using Statcast data from the past three seasons, I constructed various pitch sequence parameters to gauge the efficacy of changing the hitter’s eye level. The first parameter involved pitches that were in the strike zone, as defined by the MLB Gameday zone. Pitches in zones 1, 2, and 3 were coded as “up,” zones 7, 8, and 9 as “down,” and 4, 5, and 6 as “middle.” All other zones were considered off the plate. I focused on pitches in the strike zone because we know hitters are more likely to swing at those pitches and generally have success when they do. The in-zone swinging strike rate over this sample was 12.1%, while 28.1% of these pitches were put into play. Batters had a .349 wOBA on pitches inside this strike zone versus a .304 wOBA outside of it. Any degradation in performance on pitches inside the zone would be a real value-add for pitchers.

With each pitch coded based on its nominal location, I created pitch sequences that indicated the pitch at hand’s vertical location and the vertical location of the pitch that preceded it. Obviously, since we are looking at two-pitch sequences, the first pitch of a plate appearance will never function as the pitch at hand in our data.

I also looked at the actual change in vertical location of two consecutive pitches, in units of feet. I thought looking at pitches in the Gameday zone was a good start, but there were two issues. The first is that the zones are somewhat arbitrary by nature; someone had to decide the procedure for binning those pitches. That’s nice because it is consistent, but how those bins were defined in the first place places arbitrary boundaries on continuous data. The second issue is that I was limiting myself to pitches inside the strike zone, though for good reason. The Gameday zone has nine difference locations for pitches inside the zone, as opposed to four for pitches outside the zone. Part of the reason I chose to only consider in-zone pitches was I could not be as certain of the relative vertical height of pitches outside the zone as inside; a pitch in zone 11 could either be up high out of the zone or off the plate middle-in to a right-handed batter.

Utilizing the numerical change in location eliminates these concerns. It is independent of how two pitches are coded in the Gameday zone and can easily be applied to all applicable pitches (i.e., those that do not start a plate appearance). For context, the strike zone is about two feet high (the generally accepted range for zone height is between 1.5 and 3.5 feet), so sequences that involve changes of much more than two feet probably involve one pitch that is way above or way below the strike zone.

Finally, I created sequences based on pitch category, not pitch type. The categories were fastball, breaking ball, and offspeed (creating sequence possibilities of fastball-fastball, fastball-breaking ball, etc). The issue I had with using specific pitch types (FF, SL, CU, etc.) was that there are more possibilities for individual pitches, which makes the possibilities for sequence combinations even larger. With the three pitch categories, there are nine possibilities for a sequence value, compared to the 64 different combinations for the eight pitch types (if you combine two-seamers and sinkers, and curveballs and knuckle curveballs). The code would be almost exactly the same (if not easier because I wouldn’t have to add the step of classifying each pitch as fastball, breaking ball, or offspeed), but the presentation of so many sequences would not fit well in a digestible table for readers.

With that clerical work out of the way, let us get to the results, first via the Gameday derived sequences.

In-Zone Pitch Height Sequences
Sequence Swing% SwStr% CS% wOBA
Up-Up 80.1 19.6 14.0 .278
Up-Middle 80.3 19.4 13.3 .284
Up-Down 79.5 18.1 13.5 .284
Down-Down 75.1 13.9 20.3 .308
Down-Middle 74.0 13.5 20.7 .294
Down-Up 73.6 12.6 20.8 .292
Middle-Down 84.4 11.7 14.7 .361
Middle-Middle 84.6 11.6 14.4 .348
Middle-Up 83.3 11.6 15.8 .357
SOURCE: Baseball Savant

When looking at the sequence, the first designation corresponds to the pitch at hand while the second corresponds to the prior pitch. For in-zone sequences, three sequence types stand out from the rest of the pack: Up-up, up-middle, and up-down, with each carrying swinging strike rates north of 18%, similar whiff rates, and wOBA allowed figures within six points of each other. Pitching up in the strike zone is still the most beneficial strategy for pitchers, assuming they have the requisite repertoire of a riding four-seamer with a good breaking ball, no matter the sequence. The next three best sequences are with pitches down in the zone and the worst are those where the pitch at hand is in the middle of the zone. Looking at this table, there seems to be no benefit to changing the hitter’s eye level; it is an ordered list of where pitchers should try to locate in the zone (up, down, and avoid the middle). Does breaking it down by pitch category change this phenomenon? The short answer is yes:

In-Zone Pitch Height Sequences by Category
Zone Sequence Pitch Sequence Swing% SwStr% CS% wOBA
Down-Down offspeed-offspeed 89.6 25.8 7.5 .263
Down-Up offspeed-offspeed 88.6 25.3 7.2 .311
Down-Middle offspeed-breaking 85.6 25.2 10.4 .289
Down-Up offspeed-breaking 85.4 23.5 11.6 .261
Down-Middle offspeed-offspeed 87.6 22.8 8.9 .290
Up-Up fastball-fastball 83.9 22.6 10.1 .274
Up-Middle fastball-fastball 83.6 22.5 9.9 .281
Down-Down offspeed-breaking 84.9 22.3 12.1 .318
Down-Down offspeed-fastball 85.2 21.0 12.1 .278
Up-Down fastball-fastball 81.4 21.0 11.2 .285
Down-Middle breaking-breaking 82.1 20.5 13.7 .294
Up-Down fastball-offspeed 86.6 20.4 7.7 .270
Down-Down breaking-breaking 81.7 20.2 14.0 .297
Up-Middle fastball-offspeed 86.6 19.4 8.1 .349
Up-Up fastball-breaking 86.6 18.9 8.2 .314
SOURCE: Baseball Savant

These were the top 15 sequences in terms of swinging strike rate. We see a mix of pitches up and down in the zone; the height of the pitches often changes within the sequence. The prevalence of sequences involving pitches up in the zone in the prior table was a function of fastballs being the most utilized pitch; in the second table, all the sequences that performed well with the fastball as the pitch at hand were those where the given location was up. The best sequences involved in terms of both wOBA allowed and swinging strike rate involved offspeed and breaking pitches, most often with sequences where the height of the two pitches varied. Binning the differences in height between pitches in sequence yields a similar trend, in that the most profitable sequences tend to be derived when the pitch at hand is lower than the previous one, which is more likely a non-fastball offering.

Height Difference of Pitches in Sequence
Height Difference Bin (ft) Swing% SwStr% wOBA
(-4 to -3] 26.8 13.9 .269
(-3 to -2] 36.2 15.2 .293
(-2 to 1] 47.2 15.3 .300
(-1 to 0] 55.5 14 .315
(0 to 1] 59.3 13.3 .330
(1 to 2] 56.2 12.4 .342
(2 to 3] 47.9 10.7 .347
(3 to 4] 38.1 8.9 .375
SOURCE: Baseball Savant

I filtered by height changes within four feet in either direction because anything more than that became so large that batters almost never offered at them. The best sequence in terms of wOBA is the most dramatic change in height downward, while the next two bins represent the best sequences if you consider swinging strike rate. When you bring in pitch categories, almost all the best sequences, by wOBA and swinging strike rate, involve breaking and offspeed pitches with the previous pitch set up above the current pitch.

Performance of Sequences by Height Difference and Pitch Category
Change in Height Bin (ft) Pitch Sequence Swing% SwStr% wOBA
(-2 to -1] breaking-breaking 43.0 22.3 .225
(-1 to 0] breaking-breaking 53.2 21.5 .241
(-1 to 0] offspeed-offspeed 56.9 20.8 .241
(-3 to-2] breaking-breaking 29.9 20.4 .211
(-2 to-1] offspeed-offspeed 45.3 20.0 .269
(0 to 1] offspeed-offspeed 66.2 20.0 .265
(-2 to -1] breaking-fastball 51.1 19.6 .252
(-2 to -1] offspeed-breaking 44.1 19.4 .279
(-2 to -1] breaking-offspeed 40.5 18.9 .246
(-1 to 0] offspeed-breaking 56.2 18.7 .309
(-3 to -2] breaking-fastball 38.8 18.4 .235
(0 to 1] breaking-breaking 60.5 18.2 .276
(0 to 1] offspeed-breaking 60.9 18.2 .282
(-3 to -2] breaking-offspeed 31.9 17.5 .32
(-3 to -2] offspeed-offspeed 34.1 17.5 .257
(-2 to -1] offspeed-fastball 53.0 17.5 .276
(-1 to 0] breaking-offspeed 54.7 17.4 .270
(-4 to -3] breaking-fastball 29.6 17.1 .202
(1 to 2] offspeed-offspeed 65.9 17.1 .310
(-1 to 0] offspeed-fastball 62.2 16.8 .298
SOURCE: Baseball Savant

The best sequences involve changing the eye level of the hitter. For fastballs you would rather climb the ladder, while for the other pitch categories you would prefer to leave your pitch even lower.

These relationships also hold no matter the count, as we see consistent swinging strike rate distributions across each number of strikes on the hitter. The peaks for the sequences involving breaking pitches and offspeed pitches occur towards the left of zero (so pitches lower than the previous) while all the fastball-led sequences peak to the right of zero (pitches higher than the previous).

Clearly pitchers are most effective at inducing swinging strikes when they alter locations up and down. I will note that these trends are broad generalizations. Even though sequencing pitches with fastballs up and secondary stuff down is most beneficial, it does not mean pitchers should be looking to employ this strategy with every pitch sequence. If that were the case, hitters (and analysts) would pick up on it quickly and sit on those pitch types in those locations.

Nevertheless, this research does lend credence to the intuition of pitchers and coaches (and the prior research I cited above) that changing the eye level of the hitter is beneficial. Hitting involves making a series of decisions at such a rapid speed that introducing a variable to the equation (in this case pitch height) makes processing a pitch and making the correct decision that much more difficult. Even if the pitcher is peppering spots in the zone that he thinks the hitter cannot do much damage to, hitters are so talented that failing to strategically shift their gaze from one pitch to the next can spell doom. There is certainly more work that can be done on this subject. I just observed this one variable for a pitch, but the next question is how does this play into pitch effectiveness relative to other factors? Is it more or less important than location, spin efficiency, or velocity? Do certain pitcher archetypes benefit more from certain sequences, such as a sinker-baller versus a high arm-slot guy with a riding four-seamer? That remains to be seen, but I think this is a good start and should give future researchers confidence that strategically varying pitch height is a vital factor in both inducing whiffs and suppressing the opposition’s overall results.





Carmen is a part-time contributor to FanGraphs. An engineer by education and trade, he spends too much of his free time thinking about baseball.

newest oldest most voted
Cameron
Member
Member
Cameron

Great article! One practical takeaway from this is that throwing a high fastball seems to decrease the probability that the batter swings at a low fastball on the next pitch, so pitchers could try to throw a called strike near the bottom of the zone right after a high fastball. This completely checks out with my experience as a hitter. It’s hard to pull the trigger on a well placed low fastball after seeing it up.