January 22, 2016
A Refresher on Changeups
A few years ago, Harry Pavlidis presented some research on what makes a good changeup (part 1, part 2). In the first part of Harry’s analysis, he identified a few key truths about changeups that I’ll include below for quick reference:
1. The faster a pitcher's fastball, the more likely he was to get whiffs with his changeup.
2. The difference in a pitcher's fastball and changeup velocity had a similar relationship.
3. Pitchers with high changeup whiff rates threw the changeup more often.
4. The vertical "drop" of the changeups relative to fastballs impacted the whiff rates that the pitch induced.
I thought that ground-ball rates would be less interesting than whiff rates, and left them out of part one, but the evidence suggests that I was wrong.
1. The mild relationship between fastball speed and changeup whiffs is retained with ground balls.
2. There is a strong relationship between changeup velocity and ground balls—higher-velocity changeups induce more grounders.
3. Changeup speed doesn't impact whiff rate, so this seems like a winning move (but see the caveat in point 5).
4. The smaller the gap between fastball and changeup speed, the more grounders a pitcher induces on changeups.
5. Increasing changeup speed but not fastball speed will hurt whiff rates, a gigantic caveat to point 3.
6. The more sink, the better—no surprise here. Also, the more the changeup sinks relative to the fastball, the greater the ground-ball rates it induces.
Let's put those together:
1. Hey, velocity is good!
2. If you throw a hard changeup (relative to the fastball), you're trading more ground balls for fewer whiffs.
3. Power pitchers (big fastball and a good gap to the changeup) get the best of both worlds (whiffs and grounders).
4. The bigger the drop, the better—sink it, and sink it more than the fastball, and you'll miss bats and get grounders.
What we’ll do now is revisit the second half of Harry’s analysis for 2015 pitchers. Specifically, we’ll look at how we can categorize pitchers across the majors by analyzing their changeups and sorting them into buckets based on a few key attributes.
The first set of charts below presents the raw data in an interactive format. There’s a lot going on, so I’ll attempt to provide a bit of a primer:
Black/Grey/Red points – These are individual pitchers, plotted by GB/BIP (y-axis) and Whiffs/Swing (x-axis). The color of each point represents the average velocity of the changeup for each pitcher, with red being the fastest, and black being the slowest.
Green line – This is a trendline for the data, showing that there’s a very weak but positive correlation between the two axes. This matches, nearly exactly, what Harry found back in 2013.
Axes – These represent the means for each axis. Specifically, 27.8 percent on the Whf/Sw (x-axis) and 50.1 percent on the GB/BIP (y-axis).
Pitchers in the top right quadrant have some of the best changeups in the game. They get a lot of groundballs, and also a lot of whiffs. Some of them throw their changeups extremely hard while others are middle of the road. It is worth noting, however, that very few soft-tossing pitchers live in this “Good” quadrant. They are:
This likely comes back to Harry’s point that “Hey, velocity is good!” What that table doesn’t necessarily tell us is how each pitcher’s changeup compares to his fastball, a key insight from Harry’s initial investigation. As a result, I’ve duplicated the chart from above, but adjusted the velocity to show the difference between each pitcher’s fastball and his changeup. Pitchers in red have large gaps between the two pitches, while pitchers in black have tighter gaps:
For the most part this is the exact same data, only with the adjusted velocities. Some of Harry’s findings are clear as day when looking at the data through this lens. For example, Harry wrote that pitchers with large gaps in velocity were more likely to generate whiffs. Every red point on the chart is in fact above average in terms of whiffs per swing.
Harry also wrote that the alternative seemed to be consistently true as well. Smaller differences (which are much more common than large ones) tend to produce groundballs at an elevated rate. This one is less clear in the 2015 data, but the majority of the smallest differentials come from pitchers with above-average groundball rates.
So who has the best changeup in baseball based on this data? Any of the following can arguably make that claim depending on your personal preference for groundballs versus strikeouts:
Personally? I’m a Cole Hamels kind of guy. Feel free to hover over each player to see where your favorite players land in the landscape of changeups.
This, of course, is only part of the story. The profile of these changeups is important as well, though obviously less important than the outcomes they create. Below is another visualization that showcases changeup movement (both horizontal and vertical) as well as velocity differential.
There are a lot of interesting takeaways from this portion of the data. For example, Scott Kazmir has arguably the most extreme changeup in baseball, as it boasts a 16 mph difference off his fastball, above-average "drop," and above-average fade. The profile or description of these changeups is only one piece of the puzzle, but it does help us better understand what opposing hitters are seeing when facing these pitchers.
That leaves just one thing left in replicating Harry’s 2013 research: naming names. Below are a few tables that showcase the pitchers who make up the good and bad quadrants of the outcomes effectiveness above. The table includes not only the effectiveness measures used in the chart (GB/BIP, Whf/Sw), but also the profile attributes (velocity differential, movement) highlighted in the previous section.
The “Good” pitchers with above-average groundball and whiff rates:
The “Bad” pitchers with below-average groundball and whiff rates:
Finally, I’ll add roughly the same disclaimer Harry used when he first presented this same data for pitchers in 2013: "Bad" is a relative term, as all of these guys are major leaguers. Furthermore, what's not even considered is how a changeup may set up something else (pitching backward) or whether some pitchers’ shaky changeups are even relevant.
Special thanks to Harry Pavlidis for his support in putting this together.