clock menu more-arrow no yes mobile

Filed under:

Generating whiffs with sliders

Brian Blanco/Getty Images

Earlier this week, I explored how the movement and velocity of sliders contribute to the amount of groundballs the pitch generates. I concluded that having a high velocity and a lot of drop on the pitch leads to a high groundball rate, while horizontal movement causes the expected groundball rate to decrease.

While I was able to identify the sliders that had the best and worst shapes for getting groundballs, this gives us a view of only a piece of the sliders' results. To try to get a broader view of each slider, I constructed a similar model, but focused on explaining and projecting whiff rates.

If you haven't read part one, I would suggest taking a quick read through it before continuing on.

As I mentioned before, sliders have the highest average swinging strike rate in baseball, at 15.2%. They also have one of the highest whiff percentages (whiff/pitches), at 31.5%, only behind the splitter. Furthermore, some pitchers have managed to get extreme whiff rates on their slider in large samples, like Orioles closer Zach Britton and Diamondbacks' pitcher Patrick Corbin, who have both recorded whiff rates upwards of 50% over their (relatively short) careers.

Clearly, the slider can generate extreme whiff rates, and can be an elite pitch.

Constructing the Model

For this regression, I used the same PITCHf/x components that were in the groundball study, which were horizontal and vertical movement, velocity, and "velocity difference", which is the difference in velocity between the slider and the pitcher's fastest pitch. PITCHf/x data is from Baseball Prospectus.

After running the regression, I found that velocity, vertical movement, and velocity difference were all significant on the 99% level. The r-squared for this regression was .204, which isn't great, but it suggests the model has some statistical legitimacy. Velocity and velocity difference were positively correlated, while vertical movement was negatively correlated. Velocity and vertical movement had the same relationship in this formula that they did in the groundball model, showing clear overlap between the two. This suggests that the most effective sliders are fast and have significant vertical movement, because this will help induce both groundballs and whiffs.

Here are graphs showing the relationship between each variable and whiff rate.

Additionally, the p-values and r-squared values for each variable are shown below.

Velocity Velocity Difference Vertical Movement
t-value 6.32 5.16 2.67
p-value 1.1 x 10^-9 4.82 x 10^-7 0.00808
r-squared 0.0266 0.0535 0.0692

Here, we can see that vertical movement has the highest r-squared of the individual components, meaning it explains more variance in whiff rate than any other variable. Velocity difference was less powerful, but still explained roughly twice as much variance than velocity.

These r-squared values support the idea that sliders can fool hitters in two ways. The vertical movement on them can cause hitters to swing "over" the pitch, or a sizeable velocity difference would cause the hitter to swing early. Ideally, a pitcher would have a slider that combined both of these methods to maximize whiff rates.

With that in mind, let's look at the expected whiff rates for pitchers in the Rays organization that have PITCHf/x data available. If you'd like to plug in data and generate your own expected whiff rates, the spreadsheet can be found here.

Rays Sliders

Player Velocity Vertical Movement Velocity Difference xWhiff%
Enny Romero 85.15 0.11 12.31 38.93%
Chris Archer 89.06 0.50 7.2 36.64%
Neil Wagner 87.04 0.96 9.06 36.28%
Alex Colome 85.43 2.11 9.8 34.41%
Andrew Bellatti 82.50 0.22 11.51 34.33%
Matt Lollis 82.02 0.16 11.53 33.78%
Jose Dominguez 82.15 2.16 11.57 32.59%
Steve Geltz 82.35 1.39 10.72 32.22%
Jim Miller 83.88 1.68 9.40 32.16%
Ronald Belisario 85.52 1.37 7.65 32.08%
Mark Sappington 86.12 2.49 7.56 31.94%
Ernesto Frieri 83.10 2.10 9.63 31.18%
Kirby Yates 85.54 2.69 7.61 31.12%
Brandon Gomes 81.31 1.13 10.71 31.04%
C.J. Riefenhauser 78.33 -1.60 11.61 30.36%
Erasmo Ramirez 84.88 3.76 7.13 28.85%

While this model can inform us about the possible effectiveness of a pitch based on it's shape, there's a clear missing element in the evaluation: the usage. We can't add this variable into the formula because of the inherent bias this introduces - if a pitcher has a pitch that gets great results, he probably would throw it more. If it's ineffective, he's probably going to throw it less. So, we have to look at this in a case-by-case situation if we want to look at more than "best shape", and want to know "best overall slider".

While some consider Chris Archer's slider to be one of the best pitches in baseball, it was 36th by xGB% and had the 58th highest expected whiff rate. However, Archer has thrown 682 sliders this season, according to Baseball Prospectus, which is second most in the league. Junichi Tazawa, for example, has a higher xGB% and xWhiff% than Archer, but Tazawa only throws his slider 5% of the time. Because Archer can throw his slider 40% of the time and still get great results, we can say that his slider is objectively better than that of someone like Tazawa.

Enny Romero's slider has graded out as elite in generating both groundballs and whiffs. With his cutter and fastball, he features three clearly defined bands of velocity, as shown in graph below. This can help keep hitters of balance, and lead to better results.

His control has been a concern in the minors, but if he can bring his walks down to league average and limit the "mistake' pitches, he has the arsenal to be successful.

Much of Erasmo Ramirez' success over the past few weeks stemmed from better results on his slider. He has gotten whiffs on 40% of opposing batters' swings this season, and has generated a high number of groundballs as well.

But, I suggested a few weeks ago that his slider was likely over performing, and his performance on the pitch should regress downward. Based on its shape, we would expect it to get below league average whiffs, and fewer whiffs than it has gotten this season.

Best and Worst Sliders

Now, let's look at the sliders that have the highest and lowest whiff rates around the league, based on pitch shape.

Player Team Velocity Vertical Movement Velocity Difference xWhiff%
Sam Dyson MIA 83.68 -3.63 13.81 41.75%
Junichi Tazawa BOS 81.94 -6.84 12.91 40.51%
Josh Ravin LAN 85.57 -0.26 12.58 40.11%
Aroldis Chapman CIN 88.33 4.27 12.30 40.10%
Blake Treinen WAS 87.14 -2.45 9.91 39.99%
Austin Adams CLE 87.29 -0.89 10.42 39.79%
Garrett Richards ANA 87.67 -3.77 8.56 39.73%
Hunter Strickland SFN 86.15 -0.87 11.39 39.64%
Felipe Rivero WAS 79.84 -5.15 15.02 39.53%
Joe Kelly BOS 86.66 -1.89 10.29 39.50%

If you recall, Garrett Richards was on the list of best sliders for groundballs as well. He has managed to get a significant velocity difference on his slider while still throwing it faster than league average.

Giants reliever Hunter Strickland has one of the highest xWhiff% in the league, and his 49.31% xGB% was above league average as well. There has been turmoil in the Giants bullpen, as manager Bruce Bochy temporarily removed closer Santiago Casilla from the role last week. If Bochy decides to make a permanent change, Strickland's elite slider and sharp fastball suggest he has the tools to succeed in the role.

Now, let's look at the sliders that generate the fewest whiffs. I removed pitchers that throw sidearm or submarine, because their delivery causes the model to project them with slight inaccuracies.

Player Team Velocity Vertical Movement Velocity Difference xWhiff%
Chris Young KCA 80.67 4.51 6.40 21.87%
Kyle Lobstein DET 82.39 3.69 4.71 22.33%
Jered Weaver ANA 77.60 0.54 7.82 22.67%
Blaine Hardy DET 84.33 5.19 4.24 23.13%
Doug Fister WAS 81.34 2.40 5.94 23.59%
Casey Janssen WAS 83.37 3.08 4.77 24.11%
Eric Stults ATL 82.62 3.83 6.68 25.26%
Brett Oberholtzer HOU 83.60 4.67 6.24 25.32%
Tsuyoshi Wada CHN 82.55 3.40 6.63 25.40%
Rob Wooten MIL 83.35 2.84 5.68 25.51%

Doug Fister started throwing his slider last year, and hasn't had good results with it. He has thrown it only 5% of the time this season, but it has drawn 18.5% whiffs, and has a groundball percentage of 12.5%. He throws five other unique pitches, according to Brooks Baseball, so it is confusing that he introduced and continues to throw the slider, as it has been ineffective.

Despite high vertical movement, Jered Weaver's lack of velocity causes his sliders to generate few whiffs. But, that hasn't always been the case  - when he threw his fastball closer to 90-91 mph, he had a larger velocity difference that boosted his expected whiff rate up closer to league average. But, now that he sits around 85 mph with his fastball, the smaller velocity difference isn't enough to make up for the low velocity of the slider.

Identifying the significant components of a pitch can alter the way that players are evaluated at many levels of baseball, and creating a model to project whiff and groundball rates can help set expectations for player performance. While this is far from a definitive study, it can ideally lead to a greater understanding about pitchers in the future.

PITCHf/x data from Baseball Prospectus and