MLB: The case for robo umps

Robo umps may arrive sooner than you think: MLB is looking at data from this summer’s Atlantic League

Major League Baseball officials are beginning to process the data gleaned from an experiment conducted in the Atlantic League these past several months. The idea is to determine whether to adopt electronic strike zone technology.

Atlantic League umpires began using the technology to call balls and strikes this summer. For the past few years, the league has operated as a sort of MLB testing lab for such concepts.

The system employs TrackMan technology that will be familiar to fans who have watched big league games these past several seasons. It relies on radar to track pitches crossing home plate. In the Atlantic League experiment, the data is combined with preprogrammed specifications detailing the height of each batter’s strike zone.

The plate umpire receives a Wifi message on each pitch.

More from Call to the Pen

• Philadelphia Phillies, ready for a stretch run, bomb St. Louis Cardinals
• Philadelphia Phillies: The 4 players on the franchise’s Mount Rushmore
• Boston Red Sox fans should be upset over Mookie Betts’ comment
• Analyzing the Boston Red Sox trade for Dave Henderson and Spike Owen
• 2023 MLB postseason likely to have a strange look without Yankees, Red Sox, Cardinals

Rick White, Atlantic League commissioner, told Michael Drapeck of CBC News that “the reception has been so strong and the interest in this by people surrounding the game and on the field is so strong that inevitably it’s going to find its way to Major League baseball.”

Research, including that based on trackMan data, has shown that major league umpires generally get about 92 percent of ball and strike calls correct, the normal variance among umpires being 1 to 2 percent. Superficially, that sounds like an exceptionally accurate performance that might cause one to wonder what all the fuss is about?

But even a modestly deeper two-step dive into the numbers illustrates the problem that’s prompting consideration of electronic umpiring.

Step 1:  Although that 92 percent average sounds impressive, the reality is that most ball and strike calls are simple enough that anybody with a rudimentary grasp of the Major League rule book could get them right. About three-quarters of pitches that are taken are either obviously outside the strike zone – in the dirt, head-high or well off the plate – or they are ‘easy’ strike calls.

Batters routinely take such obvious strikes when they are delivered on first pitches, or in 3-0 counts.

Realistically, of the 150 pitches umpires call during an average game, about 90 fit the description of ‘easy calls.’ So praising umpires for getting 92 percent of calls right is akin to praising a math student for scoring 92 on a test in which most of the questions are “how much is 2 + 2?”

The real test of umpire quality comes on the close calls, those 30 to 40 decisions per game that hang on the fringes of the strike zone. Data has fairly consistently shown that umpires get about three-quarters of those close calls right…but that also means they miss about 15 per game.

Coincidentally, the strike zone’s fringes are also where pitchers tend to target count-sensitive pitches. Which leads to Step 2.

Step 2. Does it matter whether an umpire gets a ball/strike call right or wrong? In a purely statistical sense, it depends on the pitcher, the hitter and especially on the count.

To illustrate, let’s look at the 2019 performance of two groups of MLB pitchers: 10 who were exceptional and 10 who were average.

Our exceptional 10 had the 10 best earned run averages in baseball this past season. For the record, that group consists of Hyun-Jin Ryu, Jacob DeGrom, Gerrit Cole, Justin Verlander, Mike Soroka, Jack Flaherty, Sonny Gray, Max Scherzer, Zack Greinke, and Clayton Kershaw.

Question: What difference did it make to those pitchers if the first pitch was called a ball versus being called a strike?

The answer: It made a noticeable difference. When that group of pitchers operated in an 0-1 count, batters hit .273 against them. If the count was 1-0, the average batting average against them rose to .345…a 72 point difference.

Those results certainly were pitcher-dependent. At 0-1, opponents only batted .141 against Ryu, while they tied into Verlander to a .348 pace.  At 1-0, Verlander held opponents to a .214 batting average …but he was the exception.

The point is that even against good pitchers, an umpire’s incorrect first-pitch call could be expected to swing count leverage substantially.

The count leverage swing was even more dramatic once the count reached one ball and one strike. When one of our group of 10 aces got a batter into a 1-2 hole, that batter averaged just a .125 batting average. Conversely, if the 1-1 pitch was called a ball, batters hit .330, even against the game’s best pitchers. That’s a swing of 205 percentage points.

And that’s just for the game’s best. Most pitchers are not Gerrit Cole or Jacob deGrom. So let’s repeat the exercise looking this time at 10 pitchers who comprised the precise middle of the 60 pitchers working enough innings to qualify for the ERA title.

Those 10 are Eduardo Rodriguez, Anibal Sanchez, Aaron Nola, Sandy Alcantara, Anthony DeSclafani, Brett Anderson, Mike Fiers, Madison Bumgarner, Zack Wheeler, and Wade Miley.

In 0-1 counts, hitters batted .327 against those pitchers, slightly better even than in 1-0 counts, when the average was .308. Perhaps surprisingly, against average pitchers, there was little count leverage to be obtained from an erroneous first-pitch call.

Ah, but the picture changed sharply on an umpire’s bad 1-1 call. Against our average pitchers, hitters batted only .152 when the count reached 1-2, but they teed off at a .343 rate on 2-1 counts, a 191 percentage point swing.

So while not all umpire misses are critical, in certain counts they can absolutely leverage a batter into or out of an advantageous situation.

MLB knows this. That’s why auto-strike technology is being seriously looked at.