Which pitchers have the most to lose if they drop a few ticks?
Introduction
As pitchers age they cannot maintain their velocity forever. Bodies break down, players lose flexibility, and injuries take their toll. Players who can maintain their velocity into their late thirties and beyond such as Randy Johnson & Nolan Ryan are the exception not the rule, as the aging curve below from Fangraphs shows.
 |
| Velocity is one of the pitcher statistics which is most affected by age. |
This loss in velocity will naturally affect some pitchers more strongly than others. Ryan Yarbrough won't be a significantly different pitcher if his "salad" gets much weaker. Meanwhile, Jordan Hicks will still be elite even if his sinker averages only 99mph as opposed to 101mph. However, there may be a class of pitchers who rely on their velocity to get results, and risk their production falling a long way with only a small decrease in pitch speed.
How can we go about measuring this? The answer lies in analysis of a pitcher's "Stuff" quality.
The graph above shows that four-seam fastballs are the most vulnerable to decreasing velocity, with an average change in pitch grade of around 7 points (Remember that the change from best in MLB to average is only 30 points).
Including vertical movement as well shows that it is those pitchers with elite rise on their fastball who will suffer most from a loss of velocity in the 93-95mph range. The brightest colors on the graph above shows the region where pitchers will be most affected.
How to measure Velocity Vulnerability?
Over the last year or two I have incrementally developed models to describe pitch quality without reference to the results that a pitcher gets. These models will be perfect to simulate what would happen if a pitcher lost velocity. The primary unit of output for these models is run value, this corresponds to the predicted change in run expectancy after the pitch compared to before. For example, a pitch which is very likely to induce a strikeout will have a negative run value as a run is less likely to score after the pitch.
These run values are then transformed in a grade for a pitch. These grades are defined on the 20-80 scouting scale, where a change of 10 represents one standard deviation across the population of MLB pitchers. The entire scale exists from 20-80, so a change of 10 is big, the range from elite to average spans 30 points on this scale.
My model of pitch Stuff takes inputs such as pitch speed, spin, movement, release point, and for offspeed/breaking pitches, the relation to the pitcher's fastball. It then converts these into a grade which represents how good the expected run value of the pitch is.
If I re-run my Stuff model, but with reduced pitch velocity, it could tell us something about which pitchers will be most affected by a decrease in their velocity. I chose a velocity decrease of 2mph, but I imagine the results would not change much with other similar values.
The Velocity Vulnerability metric (or VV) is then defined as the change in pitch grade between the original pitch, and the same pitch but 2mph slower.
What Makes a Pitcher Vulnerable to Loss of Velocity
Before looking at how specific pitchers are affected we should look by pitch type and see if there are any types of arsenal which will have the most change.
 |
Then the next set of pitches are are sinkers and breaking balls, with an average velocity vulnerability of around three points. Sinkers don't aim to miss bats as much as four-seam fastballs so it makes sense that they are not as affected by losing speed.
Offspeed pitches are not affected much by a drop in velocity, these often rely on deception and a velocity gap to the fastball so even if the fastball gets slower the hitter's timing can still be upset by a changeup.
Cutters are a unique case where sometimes they are used as a primary fastball, and at other times they are used as a breaking ball / hard slider. I believe this is why cutters get a negative Velocity Vulnerability, as the model thinks that the primary fastball type become more of an offspeed pitch. This is clearly unrealistic and therefore for pitchers who throw a cutter as their primary fastball I would not use these results to project what may happen as they lose velocity.
If we focus on four-seam fastballs specifically we can see that there is a peak in velocity vulnerability from 93-95 mph. It is pitchers who throw at this speed who have the most to lose from dropping a few ticks on the radar gun.
Which Pitchers have the Highest Velocity Vulnerability?
If we add up the Velocity Vulnerability of all the pitches that a pitcher throws, weighted by how frequently they throw them, then we can produce a Velocity Vulnerability by pitcher. This will tell us which pitchers are most and least affected by losing speed as they age.
Here is a table of the starting pitchers who had the highest VV based on their pitches in the 2021 season.
| Pitcher | Velocity Vulnerability |
|---|---|
| Rich Hill | 7.5 |
| Logan Gilbert | 6.8 |
| John Means | 6.6 |
| Blake Snell | 6.3 |
| Carlos Rodon | 6.0 |
Here is a table of the starting pitchers who had the lowest VV.
| Pitcher | Velocity Vulnerability |
|---|---|
| Dallas Keuchel | -1.5 |
| Hyun Jin Ryu | -1.2 |
| Zach Davies | -0.2 |
| Kyle Hendricks | -0.1 |
| Ryan Yarbrough | -0.1 |
An interesting case which is not from 2021 is that of Justin Verlander. Using his 2019 pitches he has a VV of 8.1, higher than any other starting pitcher!
A full list of VV for pitchers who threw at least 100 pitches in the 2021 season can be downloaded below:
Conclusions
Velocity Vulnerability can be a useful statistic for ranking which pitchers have stuff that will be most affected by a drop in pitch speed. This could inform drafting and rostering decisions in fantasy baseball, letting you know who may carry the highest risk.
There is the caveat that some pitchers may also be able to maintain effectiveness after losing speed due to the quality of their command. An example would be Clayton Kershaw. This is something which I have not attempted to measure here but could be a significant effect.
Appendix:
This statistic was inspired by a tweet from Rudy Gamble
Super interesting analysis. One thing that might be interesting to see is how a dip in velo might affect other features of the pitch that are used in the stuff model.
ReplyDeleteI don't think a 2 MPH decrease in velo would leave things like spin rate and movement unchanged which is what I think you assume for simplicity here. One solution might be to take a sample of pitches from a pitcher that is on average 2 MPH lover than his normal velo and take the average measurements of spin rate, movement etc. from that sample to get a better idea of how the stuff an actual pitch that is 2 MPH lower than average would be affected.
It is true that other pitch features are affected by a loss in velocity. In my preliminary analysis there were definitely reductions in spin rate and movement that were correlated with a drop in velocity.
DeleteI didn't include that here, but I agree that a more detailed analysis would include those changes. I think the changes to spin and movement were quite small (only a few %) so they're unlikely to drastically change the results.
Great stuff. https://twitter.com/rudygamble/status/1503100652607709185
ReplyDelete