In 2020, Devin Williams took the league by storm. En route to posting a 0.33 ERA, Williams struck out 53% of batters he faced earning himself a spot in rookie of the year conversations. Fueling his rapid ascent was his screwball-like changeup, but how can we analyze what makes it elite?
Plenty has been written about the uniqueness of this pitch, so here is the short version: Most changeups minimize spin rate to add a “fading” effect, but William’s changeup averaged 2852 rpm – higher than most curveballs. Paired with a good spin axis, the pitch moved a lot.
Baseball savant’s ‘movement vs average’ metrics capture this well by comparing the pitch to those thrown at a similar speed and arm slot. These metrics output that Williams’ changeup runs 4.7 inches more horizontally and dives 8.3 inches more vertically than an average changeup. Naturally the cells indicating this information are a bright cherry red to visually convey they are well above average.
I think this encapsulates much of what makes the pitch elite – it moves FAR more than most changeups. That being said, there are different kinds of changeups that are successful for different reasons. Take the changeup of Lucas Giolito. It does not viciously dive like Williams’ changeup, but rather slowly fades.
With a large fastball speed gap and excellent arm speed deception, it gets whiffs all over the strike zone. The lack of movement on Giolito’s changeup is actually fundamental to its success. It not diving away makes it hold the fastball’s plane longer and get more swings and misses.
Both the ‘Vertical Movement vs Avg’ and ‘Horizontal Movement vs Avg’ cells are shaded blue because they are both two inches below average. Giolito’s changeup drops two inches less and stays two inches straighter than an average changeup – hence the changeup subcategorization name “straight changeup”.
For each pitch type, Baseball Savant basically deems one direction good and another bad. With changeups, it rewards those that drop and run a lot to the arm side with red coloring while branding the straighter changeups with negative scores and blue cells.
Of course, they are not explicitly saying one is better than the other, but it can sure seem that way. Giolito’s fastball also drops two inches less than average, but the cell is colored red because less drop on four seam fastballs is generally good. In this nuance, these metrics can be misleading, especially for highly context dependent pitches like changeups.
Giolito’s changeup is built off of his fastball to optimize the pairing. If someone like Luis Castillo – another changeup virtuoso with a more horizontally oriented fastball- adopted the straight changeup, he would see less success with it than Giolito. It is not always as simple as ‘more drop/run is good’, but with savant’s advanced movement metrics can sure seem like it.
One pitcher actually featured both a fast, diving changeup and a slow, floating changeup: the king of pitching weirdness, Zack Greinke. Along with his other antics this year, he introduced a pitch to his arsenal that broke Baseball Savant’s pitch classification system. Hard changeups – like Greinke’s original – generally yield high groundball rates while high speed gap changeups yield more swings and misses; perhaps he was targeting the latter.
Regardless of the intention, Greinke’s attempt at a straight changeup struggled. Small sample size and command certainly played a part in exaggerating the difference, but the message from 2020 was clear: Pursuing a larger speed gap did not work for Greinke. His slow, straight changeup got obliterated without the benefit of inducing more whiffs. It might have worked in someone else’s arsenal, but apparently not in Greinke’s.
Used correctly, Savant’s movement vs average metrics are a quick way of analyzing the shape of a pitch. It is meant to be one of many tools used to analyze any given pitch. Giolito’s negatives just mean the pitch is straight, not bad. Looking at a changeup, fastball relative speed/movement, tunneling metrics, and deception would be fundamental in analyzing the pitch’s true contextual quality.
Deception is a term I generally dislike as I think some use it as a blanket term to explain what we do not yet understand. That being said, it clearly plays a huge role in changeups. Giolito is only able to get away with a 13 mph speed gap because his changeup and fastball are so hard to differentiate.
Altogether ignoring the role deception plays can lead to some wro… interesting conclusions. For example, qopbaseball’s QOPA claims to isolate pitch quality, but leaves deception out of the picture. The metric suspiciously crowns Blake Snell, Ross Stripling, and Kyle Gibson as the off-speed kings of the MLB. It also questionably puts the highly regarded changeups of Zach Davies, Johnny Cueto, and Dallas Keuchel well below average.
With the upgrades to technology around the league, we can better understand the black box of deception. Statcast has allowed for the thoughtful dissection of release points, but the transition to Hawk-Eye should allow for further improvements. With the tracking of full body movement, attributes like arm speed and mechanics variation can finally be quantified and their effects measured.
Savant’s metrics are not perfect, but neither is any pitch evaluation metric. Every metric makes assumptions and understanding those assumptions is fundamental. Using as much information as possible and weighing it properly is the best you can do. That is why teams invest so much in technology – they seek more information to better judge players. The public may not have access to full body tracking and mechanics data, but learning how to effectively use the information we do have gives us a more accurate paradigm of the world around us.