Technology promises to improve the baseball training landscape on many fronts – in scouting, player development, and even logistical/organizational ways, too. Many of the new products rely on cutting-edge technology, often untested in the field. In a former life, I served as a validation technician and engineer for a variety of technologies used by startups and large companies alike, so bringing that experience to Driveline Baseball has been invaluable when validating and testing new gear we consider using for our athletes and trainers. It’s important for our trainees to trust that we’ve validated both research studies and technology we expect to deploy internally.
There is perhaps no more exciting technology in the baseball market today than Trackman radar.
Beyond the Box Score had a great introduction to what Trackman radar actually is, when they took a look at it at Saberseminar 2014:
Originally developed to track golf swings, TrackMan has been offering their half of the StatCast system (or, at least, something similar) to organizations since 2008. Since then, TrackMan’s radar systems have shown up at amateur showcases such as the recent Perfect Game All-American event, where scouts in the ballpark were treated to real-time outputs via a web app.
How It Works: Originally developed to track golf swings, TrackMan’s baseball system uses a 3D Doppler radar system collecting 20,000 samples per second. A demonstration at Saberseminar (featuring a left-handed catcher) showed that the system reports pitch speed, revolutions per minute, spin axis (as translated to a clock face), and extension at release point. But Baseball America reports a much larger list of measurements.
How To Get It: By and large, the data aren’t available for public consumption yet. But if you’re an organization with around $30,000 to spend (maybe a minor-league team or a Division I NCAA program), you can purchase your own and use it to improve your pitchers’ breaking balls. Here’s an article that explains how different spin rates can produce sinking or rising fastballs, for instance.
Potential Use: The scouting and coaching aspects are important, sure, but that doesn’t mean there aren’t other uses….
Today, you can check out limited Trackman data (such as spin rate) on the MLB Baseball Savant Statcast Search website. However, raw data sets are still not easy to get your hands on, and the radar itself isn’t cheap (as pointed out above). Driveline Baseball has a Trackman radar installed in our pitching lab, generously on loan from Trevor Bauer. It took many months to get up and running, but with the help of then-employee Kirby Young, it’s been running for several months without much of a hitch.
Trackman told us directly that velocity and spin rate would work, but the other metrics might be slightly off due to an indoors installation. I followed up with several experts in the radar/target tracking industry with experience with the technology, and they helped guide my expectations and future research.
To date, we have validated several metrics of the Trackman radar. I’d like to take you through the fun process of how we tested our setup, and what actual “sports science” tends to look like!
Validating Trackman Spin Rate
While Trackman themselves ensured us that the spin rate would be accurate, we wanted to test this ourselves just to be sure. There are two ways to validate the data:
- Compare our Trackman spin rate data with Trackman MLB data
- Compare Trackman spin rate data with other known internal measures of spin rate
The first approach relies on having enough big league pitchers (or pitchers who have minor league Trackman data they can share) in the facility to throw off our mound. A snag in this plan was that our Trackman radar was not installed until after the professional training off-season period in 2016, so we didn’t have a large group of pro arms to test. However, after several months of rehab projects and others who provided minor league Trackman data, we were able to meaningfully compare our internal data vs. external data, which led to validating Trevor Bauer’s “Bauer Unit” theory:
Bauer Units = Spin Rate in RPM / Velocity in MPH
A pitcher’s Bauer Units measurement – for their fastball, at least – stayed consistent over the velocity range of that pitch. Put simply, if a pitcher threw fastballs from 70 MPH to 92 MPH in 1 MPH increments, the ratio of spin rate to velocity stayed the same no matter how hard the pitch was thrown.
Since pitchers don’t generally throw as hard off an indoor training mound as they do in a game (adrenaline, spikes, culture, etc), to compare spin rate from the lab to the game requires scaling up the spin rate. Bauer Units provided this scaling factor for us and worked very well.
By getting data from the aforementioned MLB Statcast Search page as well as through some internal connections, we were able to say with reasonable confidence that our Trackman setup was generating similar spin rate data to what is seen on the field in the big leagues and select minor league stadiums.
We also decided to take the 2nd step and validate the data internally using two technologies:
- High-resolution, high frame rate, professional video
- Inertial measurement unit (IMU) installed in a baseball
The first approach requires fairly expensive and high-quality camera gear. Fortunately, our Edgertronic camera is up to the task, producing videos like this overspin curveball demonstrated by Trevor and written about by Eno Sarris in his article: Let’s Invent a New Pitch.
By selectively coloring baseballs and counting the number of rotations at very high frame rates, you can get a very good approximation for not only the spin rate of the baseball, but the spin axis as well. Collecting video of pitchers throwing fastballs on the Edgertronic and comparing it to Trackman data, we found a very close match, well within the measurement error of the video camera.
The second approach requires baseballs that have inertial measurement units (IMU) installed in them. There are a few products on the market that theoretically can achieve this, though none have validation studies of their own published. We decided to use a unit from a company that we’ll keep private. We validated the IMU baseball using high-speed video much like we did with Trackman, and got a margin of error of spin rate measurement of less than 2% over several trials, including non-fastball pitches.
We then used the IMU baseball to test Trackman’s ability to measure the spin rate of the ball. The results were extremely solid:
The differences were negligible and well within standard error (especially considering they were higher AND lower), meaning both the IMU baseball and the Trackman radar were producing valid results.
Conclusion: We have a working Trackman as it pertains to spin rate!
Validating Trackman Extension, Side Release, and Release Height
While Trackman made no guarantees of their unit being able to measure other metrics, such as extension, side release, and release height variables, we wanted to see if these metrics could be calibrated or adjusted with an offset value to make them correct. So, we did what anyone else would do in our situation – we bought a “Black Flame” pitching machine as a consistent measure of release speed and location.
You’ll also see a whiteboard with lines drawn it every foot. This allows us to mark the exact release point of the pitching machine to see precisely where in front of the rubber the ball is being released. (This is done by noting where on the board it is released, then marking the ground and drawing a line to the pitching machine – if the board was directly next to the machine, you could use the board itself, but since it is in the background, the ball looks abnormally large, evidenced by the animated GIF below.)
By doing this, we can get the true “extension” of the pitch, compare it to what Trackman is reporting, and see if we can calculate an offset value – if one is even needed. The big issue here is that the Trackman radar may potentially be inaccurate over a series of data, meaning it reports wildly different extension values per pitch due to radar reflection problems (a reported issue from other experts). A secondary issue is that The Black Flame may not be as consistent as we hoped either.
Fortunately, the Black Flame was up to the challenge. It was remarkably consistent when analyzing the motion using 3000 frames per second – playing video back at 100 times slower than real-time:
With the pitching machine dialed in, we fired off ten fastballs from the mound with the following “true” measurements:
- Side Release: 0 inches (dead center)
- Release Height: 3 feet, 11 inches (3.917 feet)
- Extension: 5 inches (0.417 feet)
We used a Stalker 2 Pro radar gun to validate the Trackman’s velocities, which were almost always within 1 MPH.
We weighed the pitching machine down to prevent it from sliding around:
And we fired off ten pitches!
The ten pitches came back with the following data:
All measurements are in feet except for the last row – and all numbers are cut to 2 decimal places, hence the rounding errors.
The first thing you will notice is that the measurements are extremely consistent. This is a testament to the durability of the Black Flame setup we used with weight plates holding it down, and the Trackman radar itself being an extremely reliable device. However, there was a calculable offset for all the measurements – release height was too high, release side was too far to the right (from the pitcher’s view), and extension was too far forward. Fortunately since the data was so consistent, all we would need is an offset to apply to our Trackman data in post-processing to get MLB-quality data (most likely; I have not validated MLB data using a grouped variance technique across stadiums – I have done this for PITCHf/x with the help of Max Marchi – but I assume Trackman does a good job there and that’s a topic for another day).
Of course, we don’t want to just test the Trackman in this configuration – we need to see if Trackman will reliably report a movement of the machine to different locations of the mound. For example, if we move the machine forward, we want to see extension get larger by that same distance. So we moved the machine in the following directions for the second test:
- Side Release: 0 inches (dead center)
- Release Height: 3 feet, 11 inches (3.917 feet) MINUS THE SLOPE OF THE MOUND – WENT UNMEASURED
- Extension: 1 foot, 5 inches (1.417 feet) INCREASE OF 1.0 FEET COMPARED TO TEST 1
Due to the Pythagorean Theorum and a sloped mound, the release height will change negatively, but we didn’t measure that. Regardless, we can see the change in the data and see if it makes sense and remains consistent. A change of 2 feet of release height would be a huge issue, a change of a few inches negatively would be expected. Let’s take a look at the data!
We expected an increase in extension of 12 inches, and we got an average increase of 14.76 inches, so there was an unexpected gain of 2.76 inches overall. This very easily could have been due to placement/measurement error by the human operators (us), or could be a quirk in how “extension” is actually measured based on the slope of the mound, which makes sense. We measured 12 inches from our original spot down a slope, which is the hypotenuse of a right triangle formed with a straight line to home plate, so it’s very reasonable that it’s actually that measurement that is being tallied.
Either way, the difference is very slight, and again, the data is remarkably consistent. Test 2 was a major success.
We then decided to move the Black Flame once again back to its original “extension” position and move it laterally – moving it 12 inches to the right (arm side for a right-handed pitcher). The measurements for this test were:
- Side Release: 12 inches to the right (arm side for right-handed pitcher)
- Release Height: 3 feet, 11 inches (3.917 feet)
- Extension: 5 inches (0.417 feet)
The data for the third and final test came back:
We saw an average increase in Side Release by 1.28 feet, which is 3.36 inches farther than expected. Again, measurement/placement error could easily account for this discrepancy, and it’s very close regardless.
Curiously, we saw an unexpected increase in Extension of 4.68 inches when we moved the pitching machine over. This again could be due to placement/measurement error, though it’s fairly large.
We feel great about the Trackman installation at Driveline Baseball! There are calculable offsets for the relevant metrics we sought to validate, and it gives us a few more follow-up tests we can run, specifically seeing how Extension changes based on lateral displacement given the results of Test #3. We also have a high-speed wheel-based pitching machine arriving soon which can reach velocities of 90+ MPH and apply varying spin to the ball, so we can start to test pro-quality fastballs and breaking balls using our Trackman radar.
It should be noted just how difficult it is to get a radar device to work in an enclosed space with noise everywhere, much less an MLB stadium, and track ballistic objects with such precision and consistency. While Trackman has some known flaws and faults in special circumstances, for training purposes, we’re very pleased with our setup, and college and professional teams can use our methodology to calibrate/correct their radar units if they are having some strange outliers.
ADDENDUM: Release Height Validated
After this blog post was published, I forgot to include that we also tested a varying release height by placing the Black Flame on a plyometric wood box. We held Extension and Side Release constant and increased the height of the machine by 2 feet, 1 3/8 inches. After a few precarious tests on the box, we collected enough data to show that Release Height does in fact change as expected, within 1 inch of our anticipated values. So Trackman’s varying Release Height measurement was also validated.