PA Software Uses Your iPhone Camera and Machine Learning To Generate High-Quality Baseball Analytics
Baseball arguably has the deepest game analytics out of any major league sport. These insights are only generated during games using expensive, specialized equipment and a team of expert baseball statisticians. It would be impossible to generate similar statistics in real-time for amateur competition or recreational play. However, one avid baseball fan has built two smartphone applications to allow recreational baseball players to generate these statistics as they practice and play.
Matthew Bowen is the founder of PA Software, a baseball video analytics software startup that brings the power and accuracy of multi-million dollar Major League Baseball analytics systems to your smartphone. Bowen got the idea for developing phone-driven baseball analysis software during his freshman year at the University of Alabama in Tuscaloosa. He was watching an Atlanta Braves game with friends when ESPN’s K-Zone popped up on screen to showcase where the ball passed through the ‘strike-box.’ Many people have railed against the K-Zone being a permanent fixture on TV screens during baseball games, but Bowen saw a problem that had yet to be solved.
During an official baseball game, these baseball statistics are generated by teams of professional statisticians either from ESPN or other third parties that are dedicated to this kind of work. But these skilled data observers and collectors aren’t immediately available during a team practice, where this kind of information would be critical in maintaining or improving an individual players performance. A team would spend hundreds, if not thousands of dollars, to purchase motion-tracking equipment and work with dedicated third parties who specialize in collecting and generating these statistics. For the average baseball player looking to improve their game, the cost for these baseball analytics was prohibitive.
A combination of Bowen’s love for the game stemming from his days as a former high school baseball player, and support from his professor in the UA’s Emerging Scholars program led him to pursue research in computer vision for sports applications. His research set the foundation for him to develop two iOS applications, Pitch Analyzer — Pitch Tracker and Hit Analyzer — Hit Tracker, for any person hitting or pitching a ball to produce high-quality data instantly through their iPhone. A person recording a friend or teammate hitting or pitching a baseball through their smartphone would then upload the video footage to the hitting or pitching software application. The application would then analyze the video through a trained neural network to determine ball’s velocity or pitch trajectory, and then report the data back to the user.
These two applications rely on CoreML, Apple’s machine learning platform, to utilize the full capability of the iPhone’s graphics processing unit to provide the computational power needed to drive their neural networks. Bowen wrote the neural network code via Keras, an open source Python neural network library, and then used Keras’s built-in converter to covert the code from Python to CoreML.
“I didn’t know how valuable the apps I had built were until other people who used them told me,” says Bowen.
All those late nights and weekends spent in his dorm coding have paid off. The app has had over 15,000 installations through the App Store (the app is only available on iPhone for now), and anywhere from two to three thousand monthly active users. Developing the application while being a full-time college student became so time consuming that Bowen brought on his older brother, Will, also a UA grad and aerospace engineer, to help build the mathematical and physical models that the application uses to calculate the pitch and ball trajectory, as well as tracking the bat also. In addition to Will, Bowen recruited Andrew White and Jacob Zarobsky, PA Software’s Chief Information Officer and Chief Technology Officer, respectively, to round out the founding team.
PA Software believes that its products have a few core advantages over their competitors such as Blast Motion, Diamond Kinetics, Rapsodo, and TrackMan. Cost is a major factor where PA Software wins; the startup’s applications are free on the App Store. The startup’s competitor’s products and services run from a few hundred to a couple thousand dollars. Second, their software either performs as well or even better than traditional sensor-based motion tracking hardware in calculating the velocity or trajectory of a pitched ball. In addition, there is zero additional cost to using one’s internal smartphone camera versus a vendor’s external hardware needed to produce the same statistics.
If a hardware product fails, a person or team would have to wait to have their equipment serviced by the company, or purchase a newer version of what they were using. With PA Software being software-based, if users were having trouble operating the startup’s applications, Bowen and his team can easily push out an update to fix the problem in a timely manner.
The team’s main challenge was not building the iOS applications, but working with current smartphone computing limitations. Phone computational power has grown exponentially since Bowen started developing these applications. Previously, phones were not able to record in 1080p at 60 frames per second, which is the video quality needed for the PA Software to produce reliable results. One would have to record on a GoPro and uploading the footage to desktop versions of the applications. Now, phones can do what was previously limited to GoPros.
Going forward, the team hopes that they can continue to upgrade their analysis software and also develop a cloud service to host recorded video and data analytics. Still, the software they have built has nabbed them top awards, such as the 2018 American Baseball Coaches Association Best of Show award.
With baseball analytics in such high demand, PA Software is well positioned to give amateur and recreational players the data they need to take their game to the next level.
This post was first published on Forbes.