Stuff and Command in College Baseball
There is a reason it’s stuff and command, not stuff versus command. The two are not mutually exclusive of each other. It’s not one, nor the other; it is a combination of both. And with the right amount of stuff and command, a college pitcher becomes successful. But what is, the ‘right’ combination?
Sometimes, analysts will get so caught up in stuff, that from the outside it appears we care about only that — stuff. This isn’t the case. The reason we care so much about stuff is not because we woke up one day deciding that was the most important variable in pitching success. It is because we have spent dozens of hours doing research; creating predictive metrics, looking at numbers over-and-over, combing through thousands of possible outcomes, looking at trends between certain measurables and results. Through all of this research we have learned what does matter.
Data-driven people who are progressive do not build narratives and find ways to make everything fit into that narrative. They do the research, find what does and does not matter — and to what extent — and go from there. If they have a belief, conduct the research, learn what is wrong, then they adapt. That is the difference between “nerds” or “data-people” and old school folks. In ten years, command might be more important than stuff, and when that day comes, the progressive people will adapt, and that is what outsiders may forget.
What is Stuff+?
Stuff+ takes the physical metrics of a pitch and assigns linear values to each of the variables (release height/side, vertical Break, velocity, etc) and then weighs them accordingly based on their correlation to whatever value is trying to be predicted. With 100 representing a pitcher with average stuff, 150 would be 50% better than average, and a 50 would be 50% worse than average. The majority — 64% to be exact — are between 90 and 110, and when you filter out Ben Joyce (our version of the Chapman velocity filter), the highest is in the 190s.
Using my own model, I’m able to quantify college stuff. For the sake of proving the model’s estimation prowess, here is the correlation between whiff rate and stuff+; which is what the model was built to predict. Note: Contact rate is whiff rate reversed. So, a 25% whiff rate would be a 75% contact rate.
While Stuff+ is not sticky, in a sample of 114 pitchers, it estimated ERA highly with a r² of 0.502. Essentially, the stuff+ did a great job of telling us that if you have bad stuff, you will not be able to miss bats, and if you have good stuff, you will. Sample sizes are playing a part in this as well, due to some pitchers only having 3–5 innings thrown. But in order to account for survivorship bias, we have to include those who have not thrown a lot.
What is Command+?
Command+ is the value of a pitchers ability to locate the baseball to areas in and outside of the zone that see drop-offs in run production/value. With 100 representing a pitcher with average command, 150 would be 50% better than average, and a 50 would be 50% worse than average. The majority — 71% to be exact — are between 90 and 110, with the highest being in the 170s.
Without knowing the intended location, there is no way to quantify command. But, knowing that the majority of intended locations are on the edge of the zone, or within, we are able to make assumptions and place values on certain locations more than others. Essentially this is quantifying location, not command. Command+ is built to predict strike rate, so similarly to stuff+ and whiff rate, here is the correlation between the two.
Now having introduced the methodology behind how we are measuring command and stuff in this instance, here is how well the two are at estimating ERA.
The predictive power of stuff in comparison to command is evident here, with the latter having an r² roughly 40% higher. Not for nothing, both variables have negative linear relationships with ERA. This means that the better the command — and stuff — the lower the ERA. So both, as we all knew, are important. But, one is five times more important. Having seen the two separately, here is how the two look in unison when estimating ERA.
When it comes to good pitchers, its very rare that they’re good at one — meaning stuff or command — and bad at the other. The intangible skillsets that help create better stuff — repeating mechanics, optimized movement, athleticism — also lead to better command.
That being said, when a pitcher is void of one or the other, those that are missing the stuff element, tend to fall behind. Whereas with stuff, many find success even without adequate command. Similarly to bat speed and bat-to-ball/swing decisions, there is no point in having good command without average or better stuff.
This theory is proven right when we breakdown college pitchers into two groups, as those who lack command but have stuff fare significantly better. But, this doesn’t answer my question I posed earlier, what is, the ‘right’ combination?
When do we begin to see no statistical gain from improving command? Since we learned earlier that there is always a negative linear relationship with stuff when estimating ERA, we do not have to find the ‘right’ amount of stuff, seen as we can assume better stuff means less runs.
The tricky part is, knowing what we know, what is the ‘right’ amount of command? And how much does that number change depending on the quality of stuff (poor, average, elite, etc). To figure this out, we’ll expand our sample size by 3x and begin placing pitchers into buckets based on their stuff. This is to see if the linear relationship between ERA and command changes when you begin looking at stuff much more closely.
We begin with those that have stuff deemed poor, with an ERA of 8.60 across the sample, no strong correlation (an R² of 0.01 is very low given the p-value) is present with regards to ERA. So, the theory of “there’s no point in having command if you don’t have stuff.” If a pitcher has poor stuff, the likelihood of him being a successful college pitcher is roughy 14%. If you have poor command, you should probably be prioritizing stuff development.
As we get closer to the mean, command starts to matter more. Pitchers with below-average command very commonly see a statistical gain in having better command. Now, due to the fact that they still lack stuff, they fall behind with a 7.01 across the sample. When we limit the sample to those with average or better command, ERA drops more than an entire run to the tune of a 5.87 rate.
Thanks to the pitcher with the worst command in college distorting the sample, those with average stuff see the largest correlation with ERA out of any grouping. Even when we remove said outlier, the sample comes out to an R² value of 0.09. For the most part, pitchers with average stuff will hover around with an average ERA, even with an outlier ability to command the baseball.
As we start getting into the guys with high-end stuff, we start seeing ERA dip below the average, and command starts becoming a second thought. While some pitchers post ERAs almost in the 10s, pitchers with similar command also run ERAs in the 1s. When it comes to pitchers with above-average stuff, command has almost no effect on performance.
Pitchers with what is considered ‘great’ stuff combine for the lowest ERA yet, as we’d expect. And again, no correlation between command and ERA is present. Just two pitchers posted ERAs above 5 in this sample, with the rest all running rates under a 5. So, for the most part, command doesn’t matter thus far as a pitchers stuff strays away from the average.
Lastly, pitchers with elite stuff ran the lowest ERA of any group, but they did have three pitchers run ERAs at a rate higher than a 5. Again, no strong correlation is found between command and ERA in this sample.
What is the ‘right’ amount?
Pitchers with stuff right around average tend to benefit from having good command the most, and pitchers with poor or elite/great stuff are rarely impacted by command as their stuff is such an outlier — in a good and bad way — that their command becomes a second-thought. There are pitchers with command+’s in the 60s and ERAs in the 1s, and unsurprisingly, they have good stuff.
If a pitcher has elite or great stuff, they should be pitching regardless of command. At the bare minimum, a pitcher has to throw at least 43% of pitches in the zone to be successful in college. Once they get above that mark, there is almost no statistical gain to throwing more pitches in the zone.
When it comes to pitchers with elite stuff, command doesn’t matter, they just have to throw enough pitches in the zone. For pitchers with either great or above-average stuff, getting above that 105 range is the goal. Pitchers with average or below-average stuff should be prioritizing stuff, but they can be effective if they have a 120 command+ or better, and even those with average or better command and average stuff tend to have some success. Lastly, pitchers with poor stuff; they shouldn’t be pitching. If they are, they should be entirely focused on improving stuff. Even in college, stuff is king, and everything else is a second fiddle to it.