The Effect of Adding a New Pitch
We go back to 2015 to locate pitchers that added a new pitch mid-season and see how their performance changed afterwards, and look at pitchers who have recently added a new pitch in 2023.
I love the daily notes and the game previews and the tweeting and all of that business - but these kinds of posts are truly my favorites to do. Coding with StatCast data and investigating something specific to learn something more about the real world in baseball.
Today, I wrote a bunch of code to look back to 2015 and find pitchers that added new pitches in the middle of the season and then to check how their overall performance changed after the new pitch was added.
This sounds like a ton of work, but with Python and Baseball Savant, I came up with all the numbers I needed in less than 20 minutes - pretty sweet. So let’s get into it.
The Parameters
I limited the study to these parameters
Pitchers that made at least 15 starts in the given season
Pitchers with at least 400 pitches thrown before and after the new pitch was added
Significant usage of that pitch after it was added (at least 90 of the new pitches thrown, I used 90 because it got us a few more data points instead of using an even 100)
It turns out, that going back even the whole way to 2015 - this has only happened 31 times. Here’s the rundown.
I compared four stats from before and after
K%
BB%
OPS allowed
xwOBA
We want to study this in terms of real-life results, so even though OPS is susceptible to a good bit of randomness, I think it’s a good one to look at. xwOBA is probably the best one to look at since that pretty much considers everything, so really that’s the most important one in my view.
Results
Overall, the pitchers saw these changes after their pitch change, on average:
K% came up an average of 1.2 points (think going from 25% to 26.2%)
BB% stayed the same on average
OPS dropped .023 points on average (think going from .750 to .727)
xwOBA dropped .015 points on average (think going from .350 to .335)
That’s pretty significant stuff. A gain of 1.2 points in K% will add about ten strikeouts to a season. An OPS drop of .023 points is about one-fourth of a standard deviation in OPS. That takes you, for example, from what Julio Rodriguez did last year (.853) down to about what Yandy Diaz did.
Now, there are some important notes on the biases here that we have to consider. Survivorship bias sneaks in here. What that means is that we are only studying data points that have made it through a trial. We took out every pitcher that added a new pitch and then quickly scrapped it. We are looking at a sample where the new pitches were working well, and therefore they stuck around long enough to get included in the study.
There’s also the bias of regression itself. If you’re messing around with your pitch mix in the middle of the season, that’s a positive sign things aren’t going very well for you, and therefore your current numbers are bad. We see a few really bloated xwOBAs in the “before” column here that likely were bound to improve even if nothing had changed.
Let’s look at some examples.
2022 George Kirby
Kirby actually added two new pitches in the middle of the season, but the most notable one was the slider which he debuted on July 26th. Prior to that start, he had a 22.6% K% and a .311 xwOBA allowed. After making that change, the numbers changed to a 26.5% K% and a .246 xwOBA. We can’t really give the credit entirely to the slider, since he was tinkering with the arsenal in other ways and just getting his bearings in the Majors, but it would seem like that pitch made a difference for him.
2022 Ranger Suarez
The lefty needed another whiff pitch early on, so he added a fastball variation with the cutter on May 9th. That helped bring his very low 14% K% up to a 21% mark after he made the change, and the xwOBA fell from .352 to .288.
2021 Tarik Skubal
Skubal was getting hammered to start the year with a .443 xwOBA allowed. You could probably just stop right there and say he was due for some big-time positive regression no matter what, but he added a changeup and saw the xwOBA fall to .333 thereafter. Meanwhile, he took his K% from 17% to 28% - a huge jump.
In the sample are some pitchers that got worse after new additions were made. 2015 Jake Odorizzi added a slider early about a month in and got much worse. 2017 Dylan Bundy was similar, adding a sinker and seeing everything get worse after.
But the overall picture shows that if you can add a new pitch that is good enough to stick around, you’ll very likely improve your overall numbers.
58% of the sample saw their K% improve by at least a point
58% of the sample saw their OPS improve significantly
58% of the sample saw their xwOBA improve significantly
Another 10% or so in each category saw basically no change in those categories, so you wouldn’t say the new pitch hurt them in any way.
This wasn’t the exact same 58% every each category, but there was a lot of overlap.
It’s also worth noting that while the biggest improvers did have new pitches that performed well on average, it wasn’t entirely so. There are some examples where the new pitch was just average or worse, and things still got better for them after it was added, but it’s obvious to say that adding a new pitch that performs well will help your overall performance.
It makes some sense that even adding a new mediocre pitch can help your other pitches since it’s another pitch for a hitter to see. The research on deep arsenals against shallow arsenals is tricky due to some survivorship bias (a two-pitch starter will make it to the Majors only if those two pitches are already very good, so we’d expect them to do well, while a guy with only two pitches that aren’t both good will probably never make it to a big league rotation), but it has been proven that having a deeper arsenal does help you get deeper into the game on average.
Conclusion
I’m pretty much underwhelmed with what I found here. I could have gotten deeper into this to learn a little bit more, but overall I don’t think there’s a ton to learn.
We definitely see a history of pitchers improving much more often than they get worse or stay the same after they add a new pitch mid-season. However, it’s a pretty rare thing to happen, and it’s massively influenced by the survivorship and expected regression biases I highlighted above.
I don’t think this should change in terms of how we play the fantasy baseball game right now, but we can say with certainty that adding a pitch is a way to give yourself a chance at immediate improved performance. The problem is, we shouldn’t have much hope that the performance will improve right after a guy adds a new pitch because it’s quite possible that the pitch won’t actually stick around - most of the time they don’t.
But if we see a guy adding a new pitch, and using it consistently for a few starts in a row with at least decent results on the pitch, we should bet on that guy in some fashion - overall it would be a winning bet that he’ll improve.
I have a list of SPs who are currently healthy and have added a new pitch in-season and then proceeded to use it at least 5% of the time ever since. But first, a paywall! Subscribe today to get the real juicy part of this post, as well as all of the other stuff I do here. These kinds of posts will be coming more and more often - and all throughout the offseason. I appreciate your subscriptions!