Friday, September 6, 2019

The 2D TV Computer Strike Zone's 3D Problem

Did you know the electronic strike zone graphic used for television broadcast is significantly faulty? In this analysis, we look at the two-dimensional computer-generated PitchCast graphic and compare it to the 3D zone umpires are trained to call.

Two Different Languages...or Dimensions: It's as if baseball is speaking two different languages. Teams and fans are exposed to the familiar flat strike zone "box" at home or in the clubhouse video room whereas umpires calling the game from behind home plate are essentially playing 4D chess...only one dimension (and one sport) removed.

We analyze the following 0-2 cutter thrown by Mariners pitcher Wade LeBlanc to Yankees batter Gary Sanchez in Seattle on August 28, called strike three by HP Umpire Jim Wolf. The computer readout logged this pitch's px value (the horizontal coordinate) at 0.98, which most sources would indicate as significantly off the plate (px value of 0.98 feet is 0.15 feet from the edge of home plate [set at 0.83], or 1.79 inches off the plate).

Yet the slow-motion replay tells a different tale. Video as follows:

Alternate Link: Baseball's 2D Strike Zone Computer Problem (CCS)

Even with the UEFL's Kulpa Rule—which provides for a one-inch margin of error—in full effect, this pitch would be balled by the computer (0.79 inches away).

Pitch Break: A view from behind home plate indicates the pitch, due to the cutter's measured horizontal break of pfx_x -0.56 (corresponding to the horizontal movement from release point to home plate of 6.72 inches, in the direction of home plate's center, measured at 40 feet)

The computer cannot see this pitch's depth.
Let's assume the distance between pitcher LeBlanc's release point and the front edge of home plate is 60 feet (an extremely conservative estimate). With a pfx_x of -0.56 for a distance of 60 feet, we now need to find the comparable break (pfx_x) for a distance of just 17 inches, the depth of home plate. That comes out to 0.013 feet, or 0.16 inches.

That means, all else equal, we'd expect the pitch to break 0.16 horizontal inches after passing the front edge of home plate. Wouldn't it be great for the 2D strike zone to take break into account and allow for this extra margin?

A Math Problem: The reason this looks to good to be true is because the 0.16 inch-figure assumes a constant, linear break...and as anyone who has on-field experience throwing, hitting, or officiating pitches with movement will tell you, most pitches acquire their break later in the trajectory toward home plate.

In other words, many pitches have "late movement" or, scientifically, experience an exponential increase in horizontal (and vertical) break as said pitches arrive at the catcher's mitt.

Sidebar: The reason for this has to do with physical properties relating to inertia and flight aerodynamics. In any event, the physics is too much to bear for what baseball wants to be a "simple" strike zone graphic that is easily digested by its consumers.

ESPN attempts to simulate the final 17 inches.
All this's not enough to do math to try and solve the two-dimensional strike zone, as that doesn't fix the problem.

ESPN's K-Zone 3D tries to fix it by simulating the depth of a strike zone, but as the engineers will tell you, K-Zone 3D also relies on MLBAM's captured computer numbers, and we already know that StatCast records this data prior to the front edge of home plate. It's even worse for pitch break, which according to the documentation is measured at a distance of y=40 feet...see the aforementioned comments about late movement.

How do we fix this electronic strike zone problem? Maybe have a human step in and put forth his or her best judgment to adjudicate the final few feet of a pitched baseball.


Post a Comment