The public world of baseball statistics is so in-depth these days that most fans can have a baseline understanding of what drives player production. The underlying foundations of ERA, WHIP, K%, and even Stuff models get modern-day baseball discourse, but there’s one that seems to miss that level of scrutiny: HR/9.
HR/9 is the main statistic for identifying how many home runs a pitcher gives up, with a relatively easy scale to follow:
Generally, a sub-1.00 HR/9 is the threshold to monitor, especially considering the ever-changing run environment from year to year. Identifying outlier HR/9 figures is straightforward: Chris Sale’s 0.46 HR/9 from last year is unlikely to remain consistent (career average 0.93), while many still view Carlos Rodon’s 1.59 HR/9 as too elevated to sustain.
Typically, a pitcher’s home run problem can be overlooked in public discussions, with the analysis reduced to simply stating, “X player has a HR issue.” The baseball community is usually pretty good at identifying what drives strikeout rate, walk rate, and contact suppression, but home run rates often elude the discussion. It’s not because people are simply ignoring it, but rather because giving up a home run is usually more random than most other aspects of pitching.
Regarding other batted ball data, everything fits under specific buckets that don’t highlight outliers significantly. The occasional flyball on a sinker beneath the zone or groundball on a high fastball simply get thrown into the percentages. A home run is sometimes a park-specific event, or a hitter just “lucked out” and did the right thing at the right time.
Should Aroldis Chapman have done anything different than 102 mph above the zone? Probably not, but that’s why even Gary Sanchez is a major league hitter.
Obviously, there are skills that can prevent home runs and HR/9 isn’t a purely random stat, but there’s a lot of noise. That’s a big reason why xFIP uses a league average home run rate in its calculation. It takes almost two seasons worth of starts for a pitcher’s HR rate to stabilize, since home runs can result from a hitter doing something well rather than a pitcher doing something poor.
I wanted to look deeper into what drives a home run problem. We know that flyballs are the types of batted balls that turn into home runs, but is there another piece to the puzzle that is getting missed?
The Data
I used pitch level data from the 2023 and 2024 seasons, compiling shape, location, and results-based data for each pitch type. All analyses below are based on pitches used at least 10% of the time to exclude outliers. Since HR/9 isn’t directly applicable (or valuable) on a pitch level, I decided to use HR/BBE as my proxy. The two have a strong correlation: a 0.81 R-squared value makes me confident that we’re not straying too far from HR/9. The top players from 2024 (min. 100 IP) appear at the top and bottom of HR/9 and HR/BBE.
Pitch-Specific Metrics
Starting with shape, I used velocity, vertical break, and horizontal break as the key factors from the pitch itself. I also looked at a pitch’s Stuff+ as an additional proxy to see if a filthy pitch prevented it from resulting in a home run.
Across all pitch categories, breaking pitches have the highest HR/BBE: they see an average of 0.058 HR/BBE, compared to 0.053 for fastballs and 0.043 for off-speed pitches. Within specific pitch types, sliders and cutters see the highest HR/BBE, while changeups, fastballs, and sinkers see the lowest HR/BBE. Since different pitch types have different utilities and movement profiles, the visuals below are broken out by pitch group.
There’s a lot of noise in this data, which unfortunately is par for the course regarding home run rate. The relationship between velocity and HR/BBE is not statistically significant for all non-fastballs, and there’s a tiny relationship between fastball velocity and HR/BBE (0.01 R-squared). While velocity has immense upside for missing bats entirely, it isn’t valuable for preventing home runs. I probably could’ve told you that alone in the Gary Sanchez home run above, but the pretty charts are fun too.
However, fastballs and off-speed pitches with more rise significantly affect their HR/BBE. Intuitively, this makes sense. Pitches with more vertical movement end up in positions where a hitter is now elevating when they make contact. Conversely, fastballs with more horizontal movement have a statically significant effect on limiting HR/BBE.
You want a fastball with low vert and a lot of run? Congrats, you’ve reinvented the sinker.
Not all pitchers can (or want to) throw a sinker, but this continues the trend in baseball: having multiple fastballs is critical for limiting hard contact. However, pitches also need whiffs, which is why sinkers aren’t dominating the sport. But the pure nastiness of a pitch isn’t dictating its HR/BBE either.
There’s only one meaningful relationship between Stuff+ and HR/BBE, and that’s with breaking pitches. It’s a minimal relationship between breaking pitches’ Stuff+ and their HR/BBE, so not groundbreaking in any particular way. Since a lot of HR/BBE is built into contact suppression, sometimes pitches that don’t have a high Stuff+ do well. We did reverse engineer why sinkers are important, and those don’t often pop in the Stuff+ model, so this makes sense.
Location is Important Too
While the only true takeaway from the pitch characteristics is that pitches with more rise lead to more home runs, location should also have an effect. Throwing pitches into a hitter’s nitro zone allows a hitter to maximize their damage. When a pitch goes into other hittable locations, it’s more likely that a good hitter can hit it hard but not necessarily over the fence.
That brings us to Mistake Rate, which is powered by PLV data (which uses velocity, shape, location, and count). Mistake Pitches are pitches with a sub-4.5 PLV grade in the strike zone, accounting for 8.5% of all pitches. Visually, mistakes show up exactly where you’d expect:
As I built the dataset with Mistake Rate, I assumed all pitch types would strongly correlate with Mistake Rate and HR/BBE. I was surprised that the only place it’s significant is with fastballs, and the opposite of what I expected was true.
Pitchers that throw less mistake fastballs see an unexpected increase in HR/BBE, but the data is just more noise for breaking and off-speed pitches. While true mistakes might not be getting punished for home runs as much as we’d think, it might just be the overall location of the pitch.
Fastballs and off-speed pitches correlate their HiLoc% (High location rate) and HR/BBE, which ties into pitches with more rise and Mistake Rate. For fastballs living at the top of the zone, if they miss their spot and are in the strike zone, it’s a perfect opportunity for a hitter to launch a home run. On the contrary, off-speed pitches than can’t be keep down can lead to damage. For example, Grayson Rodriguez struggles to throw his changeup below the strike zone despite still being an effective whiff pitch. Even though it can get whiffs, it’s not this magical “never gets hit” pitch. He’ll leave the ball up in the strike zone, and if a hitter is prepared, it will lead to more home runs than expected.
Continuing the use of PLV data, we also have the predicted HR/BBE that we can compare to actual HR/BBE.
This is the most decisive data, with the best relationship for fastballs (0.21 R-squared) and off-speed (.12 R-squared) yet. However, breaking pitches still don’t have a statistically significant relationship. The predicted HR/BBE also has an inverse relationship with Mistake Rate, which is fascinating. More mistake fastballs lead to more batted balls, but not home runs.
With the predicted HR/BBE, we can look at the individual standouts on both ends of the spectrum to see if there’s a reason for these outliers.
Reynaldo Lopez’s fastball is the biggest overperformer in HR/BBE vs. predicted, it had a 0.010 HR/BBE while PLV predicted a 0.051. Lopez threw the fastball 55% of the time last year and only came away with a 2.9% HR/FB% (94th percentile). The pitch itself is good but not great, and this seems like somewhere Lopez could have a harsh overcorrection in this season. Most of the other overperformers were fastballs, but some other pitch types did enter the mix.
Further up the list included pitches from Chris Sale and Seth Lugo, two pitchers who had unexpectedly exceptional seasons too.
Meanwhile, Paul Skenes‘ slider was the biggest underperformer in this analysis, with a 0.125 HR/BBE compared to a predicted 0.030. The slider was a tool that he used to prevent hard contact (23.1% ICR) and not necessarily earn whiffs, but it was susceptible to flyballs. The volatility of flyballs turned into a handful of home runs on early count sliders, which appears to be bad luck. With the underperformers, there are mostly breaking pitches.
Interestingly, Colin Rea’s fastball was one of the most overperforming pitches while his cutter was one of the most underperforming. Since cutters are often used for contact suppression, this seems backwards. Rea’s fastball grades out poorly and the cutter is slightly better, suggesting that the cutter masks the fastball better than it acts as a cutter itself. Rea also has a sinker, giving him three fastballs to deploy to hitters.
Location Leads to Contact
When pitchers keep pitches up in the zone, mainly the heart of the zone, it allows hitters to create more flyballs than expected. Outside of line drives, flyballs are the worst contact to give up: they have a .420 xwOBA compared to a .232 xwOBA on groundballs. This type of contact finds gaps deep in the alley, and you know it, leads to home runs.
While this is the most straightforward graph, all pitch types have a statistically significant relationship. Having flyball tendencies lends itself to seeing those become home runs, even if a pitcher is good at suppressing hard contact.
Searching for Breaking Ball Answers
The answer for fastballs and offspeed pitches is generally to avoid being in the upper part of the zone with pitches that rise (i.e., make sure it’s getting a whiff). There hasn’t been an explicit identifier for breaking pitches yet though. There’s a small relationship between Stuff+ and HR/BBE, but it’s not enough to necessarily be a takeaway. Even if a breaking pitch is filthy or isn’t up in the zone frequently, it can still result in home runs.
I looked at different locations, but nothing yielded meaningful data for what leads to breaking ball home runs. John Foley previously wrote about the increase in high breaking balls and how it can be a productive tool for pitchers.
Is this all to say that it’s purely luck? I don’t think so. I believe the results of breaking pitches in the zone depend more on sequencing, but I don’t have the data to back that up in this piece. A lot of breaking pitches in the zone earn called strikes, which opens the door for a hitter. If a hitter jumps on a “get-me-over” curveball on 0-0, sometimes nothing can be done except sequencing.
Conclusions
Pitcher home run totals are a fascinating beast, since no single direct metric shows how to suppress home runs (outside of pitching in a nice home ballpark). Limiting flyballs is always a good way to limit the potential for home runs, but that comes with limitations.
Pitchers want strikeouts, but often have to earn whiffs above the strike zone with fastballs or around the zone with secondary pitches, leaving pitches susceptible to flyball contact. While having a sinker in addition to a fastball helps limit home runs, that’s often easier said than done. Breaking pitches also act vastly differently than the rest of the group: their less vertical movement profiles make them harder to understand when it comes to home runs.
There’s no special sauce for avoiding home run problems, but if a pitcher is doing the small things right, they can ideally mitigate the chance of running into a bad spell of home runs.