Notice: Trying to get property 'display_name' of non-object in /var/www/html/wp-content/plugins/wordpress-seo/src/generators/schema/article.php on line 52
keyboard_arrow_uptop

What makes a good changeup? Speed differential (i.e. being 7-10 MPH off a fastball)? Is it depth or fade, perhaps the tumble on a splitter? Location and command? Deception (e.g. matching arm speed and release point)? Or is it context, how the batters are setup based on the count or the read of their swing?

Oh, right. Results. We judge on results.

I'd like to think it's the other factors that lead to the results.

Casting the net
Results are what matter most, at least at this level. "Level" in this case is the major leagues dating back to 2011. Using pitchers who threw at least 300 changes (including splitters) and at least 1,000 fastballs (including sinkers but not cutters), I went looking for case studies to peruse. I'm hoping to find pitchers who "should" have good changeups but don't.

For assessing the quality of result I used two of my favorite metrics:

  • Whiff (misses per swing)
  • GB rate (grounders per ball in play)

Then it was time to throw a bunch of other metrics into mix: velocity, movement, release point and swing rate went into the kettle, too. Chase (swings outside of the zone) and Watch (takes inside the zone) were added for additional seasoning and to bind the other ingredients.

The final ingredient was a actually trio of related others I cooked up out of convenience. Using the "zone location" parameters you have seen in our Pitcher Profiles, I took center (three squares across the middle) and added "up" (three top squares in the zone and one middle-up out of the zone) and "down" (three bottom squares and one middle-low).

Due to a combination of lack of time and a desire not to slice up my sample too far, I did not split by batter and pitcher hand. That's for the deeper perusal. I also used Excel and correlations, so I expect hate mail tweets. I broke down metrics for changeups and fastballs, and then calculated the differences in those metrics between fastballs and changeups and looked for relationships that had some power.

I didn't even get to ground balls. There was enough intrigue in the whiff data alone to get my brain swirling.

Want to miss bats with your changeup? Throw fastballs with a lot of velocity have some depth on your change relative relative to your fastballs. The best changeup whiff rates since 2011 belong to Stephen Strasburg (.54 rate, 96.4 mph fastball) and Tim Collins (.53, 93.5). Eight of the top 10 throw fastballs averaging better than 93 mph, with only Cole Hamels and Francisco Rodriguez being exceptions. The bottom 10 has only two guys over 93–well, 92.8 to fit Mike Pelfrey (.16, 92.8) in with Derek Holland (.15, 94.3).

The overall correlation between fastball velocity and changeup whiff rate resulted in an r2 of just .076. That's almost identical to the correlation between whiffs and the speed gap (fastball mph – changeup mph). Both correlations were positive and less powerful than the negative correlation with changeup whiff rate and fastball/changeup ratio.

The more heavily a pitcher relied on fastballs, the less likely he was to miss bats with his changeup. This is called "selection bias". Guys with crappy changeups tend to know it and not throw it. The exceptions are intriguing.

The typical ratio is somewhere around 4.5–so say two changes for every 10 fastballs. Roughly speaking. Strasburg is the only pitcher in the top 19 who has a ratio over 4 (4.1). Josh Tomlin resides in the bottom 10 but comes in with a 3.6 ratio. Jair Jurrjens is next to last but has a tiny 2.3 ratio. Vance Worley and Aaron Harang bracket Jurrjens in the bottom three but both have ratios over 10. Smart fellas.

The r2 for the changeup whiff to ratio was .198 (r=-.45), which was nearly matched by one measure: the difference in vertical movement (pfx_z) between the changeup and fastball. Using the drag-removed/gravity-added numbers I feature at Brooks Baseball, came in with r2 of .164 and a positive relationship.

The last relationship, expressed in a pair of correlations, is the likelihood of a swing at all as well as the rate of changeups chased out of the zone. Changeup swing rate by itself has similar predictive power as ratio and vertical movement gap (r2=.166) but even more important is the Chase rate (r2=.274). This speaks to deception and settting up batters, getting them to expand the zone and/or fail to recognize the changeup.

There are numerous weaknesses with this analysis. There are other things to measure and multivariate analysis is called for. Ground balls need some love. That will be Part 2. By the time we arrive at Part 3 we'll go outside of PITCHf/x and baseball stats and look at video.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
dethwurm
5/10
This is awesome! Really looking forward to the subsequent parts.

Any plans to do the other pitches?

(Also, perhaps I'm in the minority, but I like having summary tables/charts presenting the relevant data right next to each other. Just a suggestion!)
harrypav
5/11
Thanks Pat. Summary data is a good idea, I will include that in the next segments.

As far as other pitches go, no current plans but I can see that happening.