Umpiring: A Look under the Hood

Times have changed for the Men in Black

Baseball churns along here in mid-August to its old and familiar rhythm.  The Division chases warm up for some and chill for others.  The number of days left on the schedule begins dwindling.  And even the most office-bound of fans notice the earlier onset of twilight.  However, on rare days baseball begs for a peek under the hood.  Some umpire will elicit howls for perceived incompetence, impertinence, or just plain belligerence. Thus it has ever been so.  However, the nature of the game has changed in a simple, yet significant way from days gone by.  Since the beginning of the game umpires were given a simple mandate: Make the call quickly, decisively, and permanently.  Whether it was correct or not was irrelevant.  That changed forever in 2014 when replay was introduced.

The ability to challenge a close play brought to light a fundamental truth: Umpires are not particularly accurate.  To understand exactly how inaccurate they are it is useful to consider the game a bit differently.  Using 2014 data the average baseball game consists of the following: There are 54 outs, 17 of those being strikeouts.  54 balls are put into play.  Of those 25 are ground balls, 19 are fly balls/ popups, and 10 are line drives.  Of that collection 16 will be hits, 2 of which are home runs.  Let’s look at that in different terms.  From the 54 balls put into play we have to get 37 outs disregarding the occasional pickoff or caught-stealing.  The fly balls will create 17 of them as 2 are homers.  Some combination of 25 ground balls and the 10 line drives will create the remaining 20 outs.  Four of the outs will come via double play.  But, before we get there realize that none of the three on-base umpires has had to make a tough call yet barring a shoe-top catch.  Assuming 8 of the line drives are  base hits that leaves us with 18 out calls to be made on the bases.  How many of them are close? Not many.

The Italian mathematician Vilfredo Pareto in 1906 developed the so-called “80/20” rule which has been adapted to all manner of situations, mostly in management circles.  In this instance 80% of all the calls a base umpire has to make are so obvious that no skill is required.  The 20% demand some higher level of skill than a person recruited off the street would have.  If we use 80/20 in the context of on-base umpiring we should have about 4 tight calls a game.  With no statistics available the eyeballs say that each manager will hold his hand up to use his decision time for a challenge about twice a game.  What we do have numbers for are challenges.  In 2016 there were 1425 challenges.  That’s roughly two challenges per three games.  The overturn rate has been steady at 49% since the system started.  Basically one call gets reversed every three games on average.   If each game averages 4 close calls, then the error rate is 1 in 12, or 8%.  What other industry tolerates an employee error rate that high?

That error rate on the bases is hardly a surprise.  It is probably higher than that by a few percentage points.  The standard on replay is not, “Correct vs Incorrect.”  Rather it is whether there is definitive evidence to overturn the original call.  Some portions of the 51% of upheld calls are also incorrect, but they lack convincing proof to overturn.  Regardless, there is now a safety net of sorts for inaccuracy.  Replay stands a fighting chance of correcting an error.  What is firmly established, however, is that umpires have an error rate.  When they put on the chest protector and start calling balls-and-strikes they carry it with them.

Two things change when an umpire goes from the bases to Home Plate.  The first is that the number of decisions to be rendered increases exponentially.  The second is that he is granted the power of Imprimatur over balls-and-strikes.  On average there will be 290 pitches in a game.  The umpire will only have to rule on roughly a third of them.  On August 8, Braves vs Nats, the Home Plate Umpire was Greg Gibson.  Per PitchFX, Gibson called 113 pitches as ball or strike.  If we apply the nominal 8% error rate, there should have been about 9 incorrect calls.  In fact, there were 10 incorrect calls per PitchFX; 8 Balls were called Strikes while 2 Strikes were called Balls.  The error rate for the night was 9%…easily within the =/- band.   But, wait, it’s worse than that in reality.

Pitchers make their living on the edges of the strike zone.  The late Kansas City reliever Dan Quisenberry famously said, “The center of the strike zone is where offensive history is made.”  Pitchers try to stay out of there.  Still, a fair amount of pitched balls find the heart of the plate.  That’s an easy call for an umpire if the ball actually makes it to the catcher’s mitt.  Way outside or a ball bounced in the dirt aren’t hard calls either.  The proverbial recruit from the bar gets those correct.  Where the umpire makes his mortgage payment is the same place the pitchers do; on the edges.  So, what if we take Gibson’s performance for the night and eliminate all the pitches except those within a ball-width either way of the edge of the strike zone.

What we get is a total of 36 pitches.  All 10 of his erroneous calls come from this collection.  That works out to close-call error rate of 28%.  On second thought, you might be better off grabbing that guy out of the bar to call balls-and-strikes.

Source: Baseball Savant

There were three fairly egregious Strike Three calls.  All of them were against the Nats.  The most obvious was to Bryce Harper.  These were the most conspicuous.  But, there was plenty of iceberg under the surface.  Outside of the numbers in the data bases the effect on subsequent at-bats is significant.  If “Blue” is going to ring you up for a pitch up near your chin, you’d be swinging at the next pitch just like it.  A few bad calls early can have dramatic effects on the game.

Baseball created a mess when it instituted replay.  Everyone knew that umpires made mistakes.  It was a part of the fabric of the game.  “Things even out,“ was the conventional wisdom.  The spectacular fails, like Jim Joyce botching Armando Galarraga’s Perfect Game and Don Denkinger’s 1985 Series-altering blown call at First Base were fairly rare.  Still, there was an aura that the umpires were pretty much always right.  The replay erased any doubt about that.  But, the leap to technology went only half way.  While left exposed for three out of every four games on the bases the umpires had their supernatural powers reinstated when assigned behind the plate.  In the meantime the home plate strike zone technology has advanced to where most television carriers have their own strike zone superimposed over the live feed.  Rather than the chest protector being the shield of invincibility it more closely resembles the King’s new clothes.  Such error rates as Greg Gibson produced on August 8 are not acceptable.

This bifurcated arrangement where technology is embraced on the bases but eschewed for balls-and-strikes is untenable.  It has been since Opening Day 2014.  It is impossible for MLB to retreat by eliminating replay.  The door to the halls of technology never shuts once it is opened.  Baseball is left with little choice but to move forward to robotic balls and strikes.  The only question on the table is, “When?”


Author’s Note: Thanks to Don Henderson for providing PitchFX data.

This entry was posted in Feature. Bookmark the permalink.