The Optimal Cutter
Model and Equation
A question I have always asked myself is what makes a great pitch? Is it the velocity, movement, or spin? Is it a combination of all these factors? I created linear regression models using velocity, movement, and spin rate of cutters as independent variables and tested it with the dependent variables of batting average, expected batting average, weighted on base average, expected WOBA, and whiff%. To determine if any of these combinations had statistical significance it is best to look at the P values of the data. If the P values of the data comeout to be under .05 then the model is statistically significant. Also if each of the independent variables used have variance inflation factors under 5 then they are not too closely correlated and the model can be used for both factors. The combination that met this criteria was velocity in miles per hour (MPH) and total movement in inches as independent variables and whiff% as a dependent variable. This combination computated as a linear regression model created the equation. .950(MPH)+ .799(TM INCHES) -84.266 = Expected Whiff %. This equation can be used to predict the expected whiff% of any cutter. For example, lets look at Corbin Burnes who threw the most cutters in the MLB last season with 1,318. Corbin had on average a 95.1 MPH cutter, and 23 inches of total break. Plugging those into the formula you get .950(95.1)+ .799(23) -84.266 = 24.456 expected whiff%. Corbins expected whiff % ended up ranking close to his real whiff% which was 27.8%. From this model it can be predicted that next season Corbins whiff% will decrease since he performed better then expect this season. This model can be used for pitchers to figure out how adding velocity and movement to their cutter will help their results. Below is a PDF with the full model analysis if you are interested in learning more.
​
​
​
​
Results
Full leaderboard can be found here-
​
​
​
Top 5 pitchers in Expected Whiff%-
1.Sonny Gray 36.75%
2.Framber Valdez 33.41 %
3.Bryan Baker 31.02%
4.Spencer Howard 30.57%
5.Yu Darvish 29.52%
​
Who is not throwing their cutter enough?
There are a few cases of pitchers who have elite cutters but have not utilized them enough. First, Sonny Gray has the best Expected Whiff% on cutters in the MLB at 36.75%. Although, he only throws cutters 9.1% of the time. Looking at the data this is one of Gray's best pitches and he should utilize it more. Gray has never really used a cutter throughout his career. He threw a cutter under 5% of the time early in his career then didn't throw it for a few seasons and now has started throwing it again in 2021. Gray may not be throwing this pitch since he does not feel comfortable with it yet or he does not know how good it is. If Gray can utilize this pitch more it can create a late career breakout. In addition, Luis Sevrino has a 28.62 Expected Whiff% on cutters. Sevrino throws cutters only 5.84%. Sevrino has a top tier cutter and is underutilizing it in his pitch set. An increase in the usage of this pitch could give Sevrino more total success. Also, Adam Ottavino should throw his cutter more. Ottavino has a 25.35 Expected Whiff% on his cutter but, he only throws his cutter 5.2%. Ottavinos Expected Whiff on his cutter is higher then his actual Whiff% on the his fastball and sinker which he throws significantly more.
Who is throwing their cutter too much?
Some pitchers throw their cutter too much and it would benefit them more to use other pitches. The first case is Jesse Chavez, he has an Expected Whiff% of 19.07 on his cutter. Chavez primarily uses his cutter throwing it around 58.5% of the time. His actual Whiff% on his cutter is around 30%. Chavez may see this and think his cutter is elite. In reality it is an average cutter and if over time his Whiff% will likely move down closer to the Expected Whiff%. Another example of a pitcher who throws their cutter too much is David Robertson. Robertson has a 16.3 Expected Whiff% on cutters but throws his cutter 50.6% of the time. The other two pitches Robertson throws have a better Whiff% then what his cutter is expected to have. Robertson should look to use the other two pitches more. Finally, Kenley Janson has a 22.7 Expected Whiff% on cutters. He throws his cutter 63.9% of the time. This is one of the highest cutters thrown percentage in the MLB. While Kenley used to have on of the best cutters in the MLB his cutter now is not as good as it used to be. He used to throw the cutter 94-95 MPH and as a result had a much higher Expected Whiff% in the model. Now he throws his cutter around 92 MPH. While Kenley still has a pretty good Expected Whiff% on cutters his cutter is regressing and it would benefit him to start using his other pitches more.
​
Outliers
Three outliers in the data are:
•Jose Alverado 55.7 Whiff % 26.73 Expected whiff %
•Huascar Brazoban 45.5 Whiff% 22.86 Expected whiff%
•Adrian Sampson 50 Whiff% 29.148 Expected Whiff%
These pitchers all have far higher Whiff% then the are expected to have on cutters. Examing each pitcher I found a reason this occurs. The three pitchers all have fastballs and sinkers that are 7-10 MPH more then their cutter. This huge differential in pitch speed can be another factor to predicting if a player will swing and miss. If a player is sitting back looking for fastball and gets a cutter that is much slower with more movement they may struggle to adjust. A future component that could be added to this model is difference between fastball/sinker speed from the cutter speed. This theory turned out to be true the other way around as well. Bryce Elder had a 6 Whiff% and a 18.41 Expected Whiff% on cutters. This was a big outlier for someone whose cutter Whiff% on cutters should be higher then it is. However, the reason this Whiff% might not be as high could be that Elders fastball and sinker are the same speed as his cutter. This means batters can have the same timing on all three pitches and the cutter is not as surprising as a pitcher who has a huge speed differential.
​
Conclusion
The Expected Whiff% model can be made for other pitches if it is statistically significant. The R square value of the model is .185 so the model covers 18.5% of outcomes. That means the other 81.5% is up to variance. Although this model is a good predictive tool there may be a model out there that provides less variance. In the future I plan to test more independent variables to see if it creates an even more percice Expected Whiff% model. Some possible variables I am interested to test are the speed differential of the cutter compared to the pitchers other pitches, the location of the cutter thrown, and the situation the cutter is thrown in. Overall, this model gives an insight to pitchers on how an increase or decrease in their speed and movement can alter their results. Pitchers can use this model in the off-season to forecast how they are developing and what results they should expect.
If there is any factors I missed or more data you would like to see feel free to reach out to me via Twitter @Greveanalytics.
​