Utilizing Stuff+ and Control+ to Optimize Pitcher Performance
This blog was authored by Connor Curtiss and Austin Marchesani
The process of pitch design has swept over all levels of baseball, and along with it has come numerous attempts at quantifying what makes a pitch good. A pitch’s value can be split up into two parts. The first is its “stuff”, which is based on pitch characteristics like velocity and movement to predict a pitch’s effectiveness. The second is its location, which we will define as the ability to generate strikes. We set out to quantify these two categories with two metrics that we call Stuff+ and Control+ to help evaluate and develop our pitchers.
Stuff+
Analysts all over the country have explored many ways to measure stuff. Some models are very complex, while others can be as simple as a linear model. Our internal model takes into account metrics like velocity, movement, and more to predict outcomes like whiffs, chases, and quality of contact. Next, it combines these predictions into an all-encompassing stuff metric.
Stuff+ is on a 100-average scale with a standard deviation of 50 at the pitch level. This means that a pitch with a Stuff+ of 200 would be 2 standard deviations above the mean, or approximately 98th percentile. The highest average Stuff+ in our dataset for any collegiate pitcher (min. 100 pitches) is 212, while the pitcher with the worst Stuff+ this season averages 16.
This metric can be leveraged to optimize a pitcher’s arsenal. For instance, a pitcher may go into a pitch design phase wanting to improve his slider. He can use Stuff+ to look into keeping the current movement profile on his slider while throwing it harder, or adding more sweep but sacrificing some velocity. Stuff+ gives the ability to test both scenarios and compare which pitch would have a higher rating.
Control+
Our location metric is much simpler than our stuff metric. Control+ measures the ability of a pitcher to throw strikes based only on the pitch type and location. A fastball low and outside will get a lower score than a slider low and outside because sliders tend to produce more strikes when located in that area. Control+ is on the same scale as Stuff+, with 100 being average. The best pitcher this season has a Control+ of 180, while the worst is 6.
On the left is an example of a pitcher with a low Control+ grade on his slider. Over half of his sliders are outside of the zone, which lowers his predicted strike rate. On the right is a pitcher with a very high Control+ score. Most of his pitches are in the strike zone, and even his misses are close to the edge. This would result in more predicted swinging strikes, which also increases his Control+.
Analysis
There are countless insights that can be derived from Stuff+ and Control+. To start, we can attempt to confirm some long-held beliefs in the baseball community. For example, we know that higher fastball velocity is typically better, but how much better? From this graph, we can see that on average, with all other characteristics the same, increasing velocity from 90 to 95 mph would increase Stuff+ from about 100 to 130. Additionally, we can observe that returns from increasing velocity are almost exponential. At higher velocities, pitchers have more to gain by increasing their velocity.
Another aspect that we have examined previously is how pitches with equal amounts of horizontal and vertical break (dead zone) perform worse than fastballs with more ride or run. Since our model takes into account the interaction between the two movement types, we can see if this relationship exists in our Stuff+ model.
The image below shows fastballs that fall near the dead zone have the lowest Stuff+ grades out of any fastballs with the same speed. Dead zone isn’t necessarily equal amounts of vertical break (VB) and horizontal break (HB), but rather fastballs that have slightly more ride than run, which is reflected in the graph. In 2022, the average vertical break on right-handed fastballs is 15.9” and the average horizontal break is 10.8”, which falls right into the center of the purple section of the graph. This best explains why the dead zone area isn’t exactly aligned with equal amounts of VB and HB.
Relationships Involving Stuff and Control
How does stuff compare with control across pitchers? Denoted by the r value of 0.108, there is a positive relationship between the two variables, albeit a weak relationship. This debunks the notion that having better stuff necessitates worse control.
Next, we can compare the correlation between overall run prevention, stuff, and control. While both variables are negatively correlated, stuff is a stronger predictor of limiting runs. There is a prominent narrative that stuff is all that matters, but it doesn’t tell the whole story because we can see that controlling your pitches is also significant to reducing runs allowed.
Putting together Stuff+ and Control+, we can get a good picture of the quality of an entire pitching staff. Here’s a plot that shows the Big Ten landscape regarding pitching. At the moment, the top four teams in the conference–Rutgers, Maryland, Illinois, and Iowa–are all in the top right quadrant of the plot, meaning they have above average stuff and control.
A pitcher’s stuff can also be useful to evaluate hitters. One problem that arises when scouting college hitters is that there are large variations in the quality of pitching that each hitter faces. A hitter with a .800 OPS in the SEC may actually be better than a hitter with a 1.100 OPS in a mid-major conference. But how could we know for sure? One way is to use Stuff+ to get a true measure of how good a hitter is, taking into account the level of competition. For example, look at these two hitters:
At first glance, the hitters look pretty similar–if anything, Hitter B has the edge. But when you break down their performance by the quality of pitching they have faced, a startling difference jumps off the page. Hitter A hits all pitching well; in fact, he hits good pitching even better than poor pitching. Hitter B, on the other hand, feasts off bad stuff and struggles against talented pitchers. This information could be used both by a hitter to help identify his weaknesses to develop further, or by a scout to project a hitter’s value when he faces professional pitching.
Conclusion
In recent years, Iowa has become one of the best places in college baseball for developing pitchers. By using a wide variety of technology along with in-depth data analysis, we are able to extract the full potential of each pitcher on our staff. Evaluating each pitcher on their ability to limit runs through complex modeling is one thing that gives Iowa pitchers an edge. Each player’s ideal arsenal is contained in certain Stuff+ and Control+ combinations that coaches attempt to unlock. That’s why having a pitching coach like Robin Lund who can understand the data, improve pitcher movements, and communicate with players effectively is so important. His expertise combined with our data analysis and use of technology makes our program so desirable for players wanting to reach their full potential.
