The Fastball Debate in Baseball
Every season has its share of articles detailing the league-wide decline in fastball usage, and 2024 is no exception. This time around, the spotlight has been on the Red Sox, who have seemingly crafted an elite rotation based on a delightfully succinct philosophy: Spin go brrrr.
Indeed, they trail the league in four-seam fastball usage by a wide margin. But they’re also ninth in sinker usage and first in cutter usage as of this writing. This is incredibly interesting, especially after you consider the graph below:
In early counts (0-0, 0-1, and 1-0), when batters are more eager to swing and hunt for fastballs, we’ve reached a new minimum for four-seam fastballs. That checks out. But look at the combined rate of sinkers and cutters: It’s back up to levels last seen in 2018. So really, the Red Sox aren’t being hipsters. If anything, they represent what the league is thinking as a whole. The uptick is there, even if you exclude Boston. The question is why?
If you look at the metrics, it isn’t so obvious. Early-count four-seam fastballs return a higher xwOBA than sinkers and cutters, but that’s always been the case. An interesting hypothesis is that for pitchers with average or below-average velocity, sinkers tend to be a better option.
Analysis of Pitch Sequencing
Today, I wanted to share my answer to the question. I didn’t set out to find an answer – initially, I was looking at ways to analyze the effect of pitch sequencing – but baseball rewards you when you least expect it.
First, I collected pitch-level data from the 2022 and 2023 seasons. After much wrangling, I was able to figure out the pitch sequence of every plate appearance. The subsequent data ranged from first-pitch popups to 15-pitch stalemates, with everything in between.
For the purposes of this project, I decided to make a few major decisions. First, I excluded plate appearances that ended in a hit-by-pitch, walk, or strikeout. Ruling out hit-by-pitches was the easiest choice: They’re the product of poor control, not sequencing. Walks have a little more to do with sequencing, since the right order of operations could coerce a batter into swinging at balls off the plate.
Second, instead of relying on literal outcomes, I calculated the expected run value of each batted ball based on the angle at which it was hit, the count, and whether the pitcher had the platoon advantage.
With the data patched up and ready to go, my model of choice was a random forest, which uses an ensemble of decision trees to make predictions.
Best First Pitches
Here’s what the data says, in order of model-predicted run value:
- Sinker: 0.0184
- Cutter: 0.0300
- Four-Seam: 0.0354
- Changeup: 0.0389
- Slider: 0.0437
- Curveball: 0.0500
Top 10 Two-Pitch Sequences
- Sinker – Four-Seam: 0.0076
- Sinker – Sinker: 0.0137
- Sinker – Cutter: 0.0196
- Sinker – Slider: 0.0205
- Sinker – Changeup: 0.0264
- Cutter – Sinker: 0.0282
- Cutter – Four-Seam: 0.0291
- Four-Seam – Four-Seam: 0.0298
- Cutter – Changeup: 0.0311
- Cutter – Cutter: 0.0330
Top 10 Three-Pitch Sequences
- Sinker – Four-Seam – Cutter: 0.0038
- Sinker – Four-Seam – Slider: 0.0051
- Sinker – Sinker – Curveball: 0.0053
- Sinker – Four-Seam – Four-Seam: 0.0057
- Sinker – Four-Seam – Sinker: 0.0085
- Sinker – Sinker – Sinker: 0.0109
- Sinker – Cutter – Curveball: 0.0151
- Sinker – Four-Seam – Changeup: 0.0156
- Sinker – Slider – Four-Seam: 0.0157
- Sinker – Slider – Four-Seam: 0.0161
