Catapult Harnesses AI To Help Solve Baseball’s Injury Problems


At a major league club’s training facility earlier this month, a coach asked his young pitchers to throw a quick bullpen session, consisting of 25 pitches off the mound. Part of the explicit instruction was “to take it easy.” A few warmed up by throwing before they stepped onto the mound and with the catcher still standing. One didn’t like his final pitch, so he threw another and another until he was satisfied.

While the coach likely assumed that these minor league pitchers had indeed thrown only 25 pitches, each was equipped with Catapult’s wearable baseball product. When the data was downloaded after the workout, it quickly became clear that a number of the pitchers had exceeded their pitch allotment, meaning the staff had a false impression of their workload.

“They were doing extra work than they were supposed to do,” Catapult’s director of business operations, Ryan Warkins, told SportTechie. “What is the difference between 25 and 41? I don’t know if we really know the answer to that, so we’re starting to try to quantify that…We’re starting to ask better questions that we can add more development to it.”

Catapult’s baseball device, which Major League Baseball approved for in-game use this season, is worn in a harness between the shoulder blades and, in addition to its GPS locator, provides data via an accelerometer, gyroscope and magnetometer to measure the lean, the turn and the force of the body. The company employs six sport scientists in the U.S., and that team has developed the machine learning algorithm to recognize patterns of exertion and identify throws, swings and sprints. One member of that team, Calvin Torres, gave an online tutorial of the potential data, including recordings of total pitches, average load per pitch, total pitch load and average time between pitches, with comparisons to each athlete’s personal benchmarks.

While Catapult is a mainstay among European soccer clubs and has been used by Super Bowl champions (2016 Denver Broncos) and NBA Finals winners (2015 and 2017 Golden State Warriors), this is its first inroads into baseball, with two big league organizations as early clients (Their identities are masked by confidentiality agreements.) Those franchises have used the same device that baseball clubs are using — the latest iteration is the OptimEye S5 — but for broader performance and position tracking information.

What the Australian wearable giant started last year with ice hockey and has expanded this year to baseball, however, is the use of artificial intelligence to identify pertinent skills.

“That’s a new thing — creating sport-specific implementations — for our device,” Warkins said, “and that’s what all that machine learning comes back to is, we have a whole bunch of raw data on these devices and what can we do from a pattern-recognition standpoint of what the motion looks like from that player and recognize events that are specific to that sport?”

Get The Latest MLB Tech News In Your Inbox!

The holy grail of all sports, especially in baseball where so much of the game has already been intensely dissected, is injury prevention. A prevailing theme is that the No. 1 indicator of risk is fatigue, yet not every athlete tires in the same way and at the same time. Pitch counts and innings limits are helpful starts but not one-size-fits-all prescriptions. Besides, not every hurler’s 25-pitch bullpen session is even 25 pitches.

“One of the challenges is that we struggle to quantify is the overall volume of work over a course of a season or over the course of a career,” Warkins said. “We can put that device on and, by the rotation of the body, we can start to measure volume — so, quantity of throws — and intensity of those throws in terms of rotation and effort put into place. We’re doing the same thing now for bat swings now as well, so how many times a guy does into a cage for any given practice.”

Warkins acknowledged that Catapult remains in the “data collection and discovery phase” whereby they are accumulating “mounds of information” about pitchers for later examination in an apropos turn of phrase he insisted wasn’t intentional. Pro organizations have at least a dozen pitchers at each minor league level, providing a robust sample size.

There are, of course, other entrants into this space already. Motus, which produces an elbow sleeve to track pitch counts and intensity, is now in its second season as an approved MLB device. Glenn Fleisig, the Ph.D. research director of the American Sports Medicine Institute, is a Motus board member who has spoken about the product’s effectiveness in analyzing elbow force. (Asked for comment about Catapult’s device, Fleisig wrote in an email that he “would like to learn more about it” but hasn’t had a chance to test the product.) Clubs also have been using TrackMan radars — the same technology used for Statcast in game situations — for years and have been known to not let a rehabbing pitcher advance until he meets certain velocity and/or spin rate milestones that match his healthy standards.

The whole research area remains in relative infancy, and Warkins said shoulder and elbow devices are “all pieces to the puzzle” and potentially “complementary technologies.” Each joint has its own reasons for careful monitoring, IoT technology may be used to coordinate between different sensors and some pitchers might find a device in one spot more restrictive of movement and less comfortable than the other.

What’s most important is developing predictors for injury, which seems to begin with tracking pitch load. Fleisig’s research has shown that the rise of Tommy John surgeries is linked with overuse of youth pitchers. Warkins said that, while Catapult is targeting pro ball first, the eventual plan is to cater to amateur markets. He noted that most leagues have instituted restrictions about how much one player can pitch and not on back-to-back days — but some play in more than one travel ball league and also play other positions.

“You’ve still got that kid playing shortstop and he’s throwing the ball all over the place,” Warkins said. “There’s some wear and tear there, too.”

Catapult’s foray into baseball should pay other dividends as well, such as measuring a runner’s explosiveness and acceleration just as Statcast does but without the large stadium-installed radars or tracking a hitter’s volume of batting-cage swings to prevent, say, an overuse injury to the oblique. Finding the formula that predicts and prevents pitcher injuries, however, will remain the ultimate objective. Research has shown that 25 percent of big league pitchers and 15 percent of minor league pitchers have undergone Tommy John surgery. Of the 400 such operations performed on big leaguers between 1974 and 2015, fully one-third had occurred in the last five seasons.

That last figure comes from an epidemiological study conducted by former Los Angeles Dodgers medical director Stan Conte, Hospital for Special Surgery orthopedist Joshua Dines, an assistant New York Mets physician, and HSS clinical fellow Christopher Camp. In considering big league injuries from 1998 through 2015, their research found that the shoulder (20.6 percent) and elbow (19.6 percent) were responsible for the most number of disabled-list stints and the greatest number of days spent on the DL: 26.2 percent and 28.2 percent, respectively.

Determining the why and the how is the hefty task that awaits.

“That’s the exciting part of this problem,” Warkins said. “Nowhere near do we have enough data to this date to even pretend to answer that question.”