Thursday, July 11, 2019

Reviewing Atlantic League's Automated Strike Zone

The Atlantic League's official automated ball/strike system debut turned in a highlight reel performance when TrackMan's electronic umpire ruled a series of pitches depicted within the graphic K-zone as balls, despite the circular baseball graphics clearly appearing in the virtual box.

Officially a debut (though MLB/ALPB tested the electronic pitch tracker throughout the season's first half), Wednesday evening's Freedom-Liberty All-Star Game in York, PA made for a slightly different type of baseball game, with HP Umpire Brian deBrauwere's ball/strike calls a little more delayed than usual, generating some gems from the broadcast booth, such as, "First pitch to Ryan Dent is a little bit...not low."

Thompson didn't like ball two.
According to one account, pitcher Daryl Thompson—who walked opposing batter Will Kengor on seven pitches, only one of which was outside of the computer-generated strike zoneactually began yelling at deBrauwere before he was informed (or reminded) that the robot umpire was to blame (Thompson then gave up a home run to ensuing batter Mike Ohlman on a 2-0 count after "ball two" was called on a pitch that the graphics indicated had half of the plate).

That means TrackMan disagreed with the computer's visual zone on four pitches over the course of two batters during Thompson's second-inning appearance, or on 4-of-6 pitches [an error/discrepancy rate of 67%].

Stats/Error Rate: Over the course of the entire game, 38 out of 101 callable pitches—nearly all of which were called balls—conflicted with the graphic depiction (37.6% error/discrepancy rate). Every inning had at least two such discrepancies, and there were several at-bats in which every pitch called during the at-bat disagreed with the graphic (visit the ALPB scoreboard page here if you'd like to verify for yourself).

The ALPB All-Star Game implementation of the automated ball/strike system (ABS) thus pitted public-facing gameday-style graphics that portrayed a series of pitches within the confines of the virtual strike zone against a robotic voice in deBrauwere's ear that continually uttered, "ball."

Visual depiction of a ball call within K-zone.
In short, red circles indicate strikes, green circles are called balls, and blue circles are balls put into play (batted balls). On more than a handful of occasions Wednesday night, a called ball appeared within or touching the virtual strike zone (recall that a strike is to be called " if any part of the ball passes through any part of the strike zone," such that a green circle that barely nicks the edge of the box is to be considered a strike).

This leads us to one of two possible conclusions: Either the graphics are incorrect or the computer has malfunctioned and failed to apply a proper algorithm or API.

HP umpire waiting to call as F2 throws to F1.
Gil's Call: My wager would be on the former: the graphics don't agree with the ABS observed values, which suggests the graphic itself is incorrect. At the ALPB level, I can't say I'm surprised. Whether it's a deficient stat stringer or some other human operator error, a computer glitch, bug, or otherwise, suffice it to say, this is not what MLB wants to see.

The league's goal is to have the computer always agree with the graphic representation given to the public, so I am somewhat surprised that of all that could go wrong, this is the problem that appeared most prominently on Wednesday in York.

Then again, as Cardiff University Professor Harry Collins stated, "technology shouldn’t be presented as showing reality, when really it’s creating reality." Collins used the tennis match replay product Hawk-Eye as an example, pointing out that Hawk-Eye simply simulates an event rather than demonstrating reality: "it’s important for the public to know the difference between what they see on their screens and what’s constructed."

The pic-error issue is fixable, but is it honest?
To fix the problem of visual discrepancy, MLBAM will have to tweak how the sport visually represents ABS; I have full confidence they can rectify the visual complication. My concern, however, is that in order to fix the graphics problem, the engineers in manipulating the graphics and/or data might get dangerously close to actually falsifying something (data, graphics, or both).

MLB Commissioner Rob Manfred has played both sides of the fence on the issue, admitting that, "that technology has a larger margin of error than we see with human umpires," while likewise acquiescing to a baseball audience hungry for a computerized strike zone: "We try to be responsive to those sorts of expressions of concern."
Related PostPodcast - Truth About Baseball's Electronic Strike Zone (6/5/19).

Strike three call after ball returned to pitcher.
Manfred also discussed the technology's development, and in doing so, revealed professional baseball's motivation for pressing on and why, as we alluded to in our Truth About Baseball's Electronic Strike Zone podcast, MLB can't afford to admit the true faulty nature of electronic pitch tracking: "We have spent a lot of time and money on the technology."

On a related note, I also wouldn't expect many complaints about TrackMan ABS for a solid year or so: fans and some players have been so passionate about getting an electronic strike zone that much like MLB, they can ill afford to complain about the tech once it debuts lest it turn into a "boy who cried wolf" scenario, especially after investing so many resources into campaigning for and implementing the computerized umpire concept.

The following video is a compilation of close/controversial ball/strike calls from 7/10/19:

Alternate Link: Highlights of Atlantic League's Electronic Strike Zone (CCS)


Post a Comment