This past summer at Driveline, we had more than 200 athletes pass through our gym, and each of them went through a full biomechanics assessment. That’s a lot of data, a lot of arm paths, a lot of lower-half mechanics, a lot of kinematic sequences, and a lot of arm stress.
From the information gathered, we’ve been able not only to figure out averages and baselines but also a better idea for target metrics that we have identified as meaningful or actionable. Knowing what averages to look for when correcting an athlete’s arm action, or what a good lead leg block looks like, is essential to correcting mechanical flaws.
We won’t go too much into how we interpret biomechanics reports for athletes, but we’ll share some observations and provide a big picture look at the data.
What Can We Learn From 300+ Biomechanics Captures?
We had more than 200 athletes and more than 300 captures, but we had to do a bit of data cleaning. Markers falling off during captures and inconsistent mechanics can lead to high standard deviations in the signals we look at. To limit variance, we removed data with more than 5% variance in torque values.
We were left with 182 captures, an average height of 6’1”, an average weight of 195 lb., and an average fastball velocity of 81.3 mph. The table below details all of the metrics that we looked at:
That’s a lot of data.
Notes and Observations
We had an average pitch velocity over 81 mph in a motion capture lab. In the past, ASMI has classified 78 mph as “elite,” so by this definition the average athlete coming through our gym is elite. That’s something to be proud of.
On the mechanics side, we now know that the average thrower at Driveline releases the ball at about 90 degrees of shoulder abduction and about 0 degrees of shoulder horizontal abduction. This means that the throwing arm is raised neutrally out to the side, similar to a t-pose.
T-pose: arms in line with the shoulders—0 degrees of horizontal abduction—and raised to 90 degrees of abduction
So, if the throwing arm is mostly neutral and in-line with the torso, what’s extension?
Extension doesn’t come from extending your arm out in front. It is actually a result of forward trunk tilt (leaning forward), lateral trunk tilt (leaning to the side), and forward shoulder rotation. While arm drag can detract from this, making adjustments with your arm is only a small piece to the puzzle.
We also know that the average thrower at Driveline keeps his trunk closed by 10 degrees at foot contact and extends his knee by 3 degrees from foot contact to ball release. But what does that really tell us?
The averages are just that. They set the baselines that we use in our biomechanics assessment, but we can’t draw many meaningful conclusions from seeing the averages of everything.
There are some valuable things though. Remember this video?
Basically, all that stuff remains true. We saw an average varus torque of 101 Nm, which is in-line with previous research. Assuming you have an average length forearm, that’s like wrenching yourself into external rotation with a 45 lb. plate.
Here’s one more fun tidbit: The average shoulder compression force, the force pushing your shoulder into its socket, was upwards of 1000 N of force. That’s nearly 225 lbs.!
Correlations to Velo
So, what actually correlates to velo?
The strongest correlations (r^2 > 0.2) are between kinematic velocities and joint kinetics. Simply put: move faster to throw harder.
We also see that throwing harder leads to higher joint kinetics, which actually contradicts past research. Basically, the harder you throw, the more stress you put on your arm.
Now the hard part will be figuring out how we can make an athlete rotate faster, move his arm faster, and extend his knee faster. We can save that for later.
Looking at kinematic positions, we actually see that no positional metrics were really correlated with velocity. Forward trunk tilt at ball release has the strongest correlation with an r^2 of 0.195. Past studies have linked forward trunk tilt to velocity, so there may be something there.
Studies have also linked maximum external rotation as well as shoulder horizontal abduction at foot contact to velocity, but we show only weak correlations.
We also see no correlation between torso angle at foot contact (meaning keeping the trunk closed), but that doesn’t mean you shouldn’t throw with your shoulders totally open. It just means we need to look at the data a little differently.
Positional metrics are harder to address in their relationship to throwing velocity. For example, someone can hit all the right positions, but just be moving very slow—and thus throw the ball slow. It’s a multivariate problem in this regard.
It could also be that the positional metrics simply don’t linearly scale with velocity. There’s a sort of diminishing returns when making changes to those metrics. For example, if you already have 50 degrees of scap retraction, increasing that even further may not make you throw harder. Or if your trunk is already closed off by -20 degrees at foot contact, chances are that closing yourself off even more won’t help all that much.
While you can hit all the perfect positions during a throw, if you’re not moving fast, you’re not going to throw hard. If you’re not strong, you’re not going to throw hard. There’s a multitude of factors that can contribute to velocity—being in the right positions is just one of them. This is where we start to see the real complexity of pitching mechanics.
Correlations to Torque
It could also be argued that some metrics imply that an athlete is throwing more efficiently—that is, with less stress.
There are definitely sequencing patterns more conducive to long term health and others more conducive to higher velo. Can I concretely say what those patterns look like/who displays them? Nope, but I can say that they exist. I think. Maybe. Who knows.
— Rob Hill (@Berticushill) January 24, 2019
Taking height and weight out of the equation, let’s look at correlations to normalized torques. I’ve provided the most interesting looking correlations, but for those interested, the full table is available.
Significant correlations are noted by the P-values with an asterisk next to them.
A couple of things make sense: There are a few correlations to kinematic velocities. The faster you move, the harder you throw, the more torque is put on your arm.
There’s also some interesting correlations to positional metrics. Lateral trunk tilt at ball release is positively correlated with varus and shoulder internal rotation torques, and shoulder abduction at ball release is negatively correlated to varus, internal rotation, and flexion torque.
Lateral trunk tilt is the amount of side tilt that an athlete has. This largely determines arm slot. Shoulder abduction is the upper arm moving in the frontal plane.
So increased lateral trunk tilt and decreased shoulder abduction at ball release both correlate with increases in varus and internal rotation torque. This is how we define “pulling off the ball”—excessive side tilt and arm dragging below the shoulder line. Here’s what that looks like in Visual3D:
We also have decent evidence that pulling off the ball has negative effects on joint torques, which is exciting.
Maximum elbow flexion has positive correlations with elbow varus and shoulder internal rotation torques. This means that being excessively “inside 90” with the arm action could have negative effects on those torques.
Elbow flexion at max external rotation is negatively correlated with elbow flexion and pronation torques, but positively correlated with shoulder adduction torque. This potentially means that decreased elbow flexion (meaning the arm is more extended) at max external rotation—what we would call forearm flyout—could lead to an increase in both elbow flexion torque and pronation torque, but an increase in adduction torque.
Front knee flexion at ball release saw significant negative correlations to varus, shoulder internal rotation, adduction, and horizontal adduction torques. This means that having a more extended knee at ball release—something that potentially indicates a strong lead leg block—is correlated to those increased torques.
This only looks at a couple of metrics available, but it provides insight defining good and bad mechanics. If we can quantify what positions actually lead to a reduction in joint kinetics, that’s immensely valuable.
Let’s look at some comparisons. First: what’s the difference between people who throw slow and people who throw hard?
We binned athletes into five groups: >87 mph, 84-87 mph, 81-84 mph, 75-78 mph, and 75- mph. Our lowest velocity recorded was 69 mph. These were all judged under a similarly-corrected alpha level using a Bonferroni correction.
Kinematic velocities are where we see some obvious differences. From the slowest group to the fastest, trunk angular velocity, elbow extension angular velocity, and lead knee-extension velocity all saw significant increases.
Simply put: rotate faster and make your lead leg block better.
Looking at the kinetics of those who threw slow and those who threw hard, we see confirmation that throwing harder results in more stress on the arm. Even normalizing for height and weight, we see significant increases in all torques except for elbow flexion torque.
Again—throwing harder is significantly correlated with increased torques on the arm.
What if we normalize for velocity to create a velocity-to-stress efficiency metric? A higher number would mean that you are more efficient—that is you generate more velocity per unit of stress. A lower number means you are getting less velocity per unit of stress.
Here we see that all efficiencies go down in the group that threw harder. The harder you throw, the less efficient your mechanics become. Although the changes were not significant (possibly due to the smaller sample size of the two groups), the results are still interesting.
What could that mean? It’s difficult to say, but perhaps as you throw harder, mechanics become less efficient mechanics. Intra-athlete testing of throwing at 50% intent all the way up to 100% intent could provide additional insight to this. This is something we’ll continue to look at as we have more high-level pitchers throw in our lab.
Now, let’s look at lefties versus righties.
Positionally, no differences stand out. From a mechanical standpoint lefties and righties move about the same.
Looking at the kinetics, we see something interesting. When normalizing for height and weight, varus torque, shoulder internal rotation, and shoulder adduction torque were all significantly lower in lefties.
Interestingly, average velocity by lefties was lower, which could have led to the reduced torques. When factoring in velocity, only varus torque velo efficiency was significantly correlated—which is still impressive. Lefties averaged 16.1 mph per unit of varus stress, whereas righties only got 13.2 mph per unit of varus stress.
What this means is that if you’re a righty and your arm hurts, you should consider throwing with your left arm. Is that how that works?
It’s worth noting that the sample size of lefties was on the lower side (n=25), so a larger sample size could provide us with more insight into these differences.
There’s a lot we can take away from this information. We’ve shown that the faster you move, the more likely you are to throw hard. The harder you throw, the higher the torques on your arm. We’ve quantitatively shown that pulling off the ball can lead to an increase in torques on the arm.
But there are still a lot of unanswered questions.
We weren’t able to correlate any positional metrics with throwing hard. Beyond kinematic velocities, we weren’t able to really see why people throw hard.
There’s also an abundance of other metrics that we haven’t looked at. Timing and sequencing obviously plays a huge role in the delivery, which we didn’t address. We didn’t look at anything in the glove arm—can we quantify positive disconnection? Those are just two ideas to start.
The metrics examined here were generated by an older version of our pipeline. In the newer version, we hope to better answer these questions and look at some other metrics that we think could be interesting.
The pitching motion is incredibly complex system that’s reliant on many factors: strength, speed, timing, flexibility, and many others. Using an ever-expanding list of biomechanical markers and metrics, integrating force plates, and focused research studies, we can continue to enhance our understanding of the pitching motion.
Down the road, initiatives like forward dynamics and more in-depth athlete typing will hopefully provide us with a more complete picture of what makes pitchers throw and what we can do train them. We’re still only at the beginning.
This article was written by Research Director Joe Marsh
The post Biomechanics Rewind: A Look at the Numbers from the Last Six Months appeared first on Driveline Baseball.