There’s Lots of Physics To Do Now That Hawk-Eye Is Up and Running

I just got done streaming the MLB Statcast Update from the 2020 SABR Analytics Conference.  Wow!  MLB’s baseball research analyst Jason Bernard, vice president of baseball data Greg Cain, baseball scientist Clay Nunnally, and director of baseball research and development Daren Willman shared the status, abilities, and potential of the new data collection system for Statcast called Hawk-Eye.

They reported that MLB has the new data collection system up and running in every park – a bright spot in an otherwise disturbingly messy offseason.  The previous measurement system was a hybrid of radar by Trackman and video technology from ChyronHego.  Hawk-Eye uses only video cameras to collect raw data and, most likely, a massive amount of state-of-the-art computing power to manipulate the flood of images into useful tools.

The new system has 12 cameras arranged around the ballpark.  Five of the cameras look at the area between the mound and home plate. You might recall the original SportVision (now SMT) system used only three.  These cameras are designed for pitch tracking as well as monitoring the pitcher, catcher, batter, and perhaps the umpire as well.  These 8-megapixel cameras collect 100 frames per second. The remaining seven cameras are 11-megapixel, 50 frames per second, and can track the motion of everyone on the field (fielders, runners, coaches, umpires and, who knows, crazed fans).  All told, the cameras generate 10 gigabits of data per second.

Preliminary measurements of strike zone accuracy indicate the position of the pitch can be detected to within, plus or minus, a fourth of an inch horizontally and even better vertically – slightly better than 2018 and 2019 results from the older method.  In regard to batted balls, where the older system was known for losing high fly balls and having to calculate the rest of the trajectory, Hawk-Eye is able to reacquire the trajectory of high fly balls that leave its field of view as the ball drops back down into camera range again.  

In addition, the previous system had an uncertainty of plus or minus 15 feet for landing distances, often due to losing the ball.  Experiments with the new system indicate a plus-or-minus one-foot error. Trackman radar struggled with balls hit down the lines because, as physics tells us, radar works best when something is moving directly at or directly away from it.  Since Hawk-Eye uses only video, it should have no such issues. The radar also was notorious for losing balls once they hit the grass. It was bad with grounders and awful with bunts. Again, the new system should be a substantial improvement.

Hawk-Eye is able to detect previously unmeasurable properties of pitches, allowing physicists to improve and advance their models for the flight of the ball.  The old system was able to use radar to measure the total spin on the ball. The effective spin (sidespin and backspin) had to be calculated based upon a model of the flight of a baseball and the ball’s trajectory.  This sometimes led to embarrassing and physically impossible results, like the occasional times when the calculated backspin plus sidespin was greater than the total spin.

In addition, Hawk-Eye is able to measure each component of the spin (sidespin, backspin, and gyrospin) independently using a methodology similar to what Rapsodo is rumored to use.  Individual video frames can resolve the seam pattern on the ball!  A series of such images not only can calculate the rate at which the ball spins but also the axis the ball is spinning around.

The accuracy of this method is such that MLB claims it has detected “spin decay” over the flight of a pitch.  Spin decay – I suppose I should explain that. If a beach ball is floating in a pool and you spin it, the water will exert friction on the ball and rather quickly stop it from spinning.  A baseball spinning in air should feel friction from the air, which should slow down the spin rate – very, very, slowly. Yet, Hawk-Eye claims to see it.

So now, instead of using a trajectory model (which is known to be incomplete) to predict the spin of a ball based upon the ball’s motion, physicists hopefully will be able to have the data to attempt the challenging science of making a much better trajectory model that more accurately incorporates spin.  After all, if you completely know the details of the motion, including the spin and spin orientation, you should be able to make a model that properly incorporates all the data properly, including the spin. Perhaps the current questions as to whether the motion of a pitch is affected due to the location of the seams during flight can be addressed. 

In addition to Hawk-Eye providing new data to advance our models of baseball trajectories, it also will make measurements that add to our understanding of the ball-bat collision.  The location where the ball collides with the bat, the speed of the bat, and the swing plane of the bat with respect to the pitch trajectory all can be detected. Presumably, the resulting exit velocity, spray angle, and launch angle also can be detected to greater accuracy.  In addition, one might suspect the launch spin and launch spin axis could be known. Knowing all the launch parameters including spin would allow physicists to extrapolate any improvement in trajectory models from pitches into trajectory models for batted balls.

Of course, MLB didn’t set-up the Hawk-Eye system just as a giant physics experiment (although it does feel that way).  There will be plenty of perks for sabermetricians, such as more data with higher accuracy and precision. The cameras have high enough resolution for 19-point skeletal tracking, so the actual body position of players will be recorded in each frame. This could be a boon for pitching coaches and hitting instructors and aid training staffs with injury prevention.  There should be plenty of upgrades for broadcasts and fans. There is even a potential that Gameday will go “back to the future” and again animate plays, but this time it will use actual position data from Hawk-Eye in nearly real time.

I guess we physicists won’t be the only ones scrambling to catch up with the huge new data stream Hawk-Eye will send flowing out once we are able to get the season under way.