Category Archives: Baseball backstop

Robo-umps continued: Applying computer vision to the job of calling balls and strikes

Back in 2015, I attended one of two historic professional baseball game between the San Rafael Pacifics and the Vallejo Admirals in which balls and strikes were determined by the advanced, multi-camera system called PitchF/X instead of a human umpire. I wrote about the game in a post and, every now and then, I’m intrigued to see headlines about potentially replacing or assisting home plate umpires with these challenging calls.

Fangraphs’s Roger Cheng published a two-part series in October 2018 on applying machine learning techniques to ball/strike calls. I look forward to reading the articles and I’ll write about them ASAP. My first reaction is that I don’t know what there is to “teach” when the PitchF/X approach has unambigously the information needed. That said, I’ll reserve judgment until I finish the articles.

Luddites beware: Umpires may be getting better, but they can’t beat PitchF/X behind the plate

My friend alerted me a recent article by Noah Davis and Michael Lopez on improvements in umpiring published at fivethirtyeight.com. I’ll comment on the article more in a moment, but reading it gave reminded me that I wanted to write about a related topic: the historical event last month in which a computerized camera system called balls and strikes in a professional baseball game. The two tie together, as you will see.

 

As reported by CSN Bay Area, an independent league team called the San Rafael Pacifics played two games using the Pitch F/X system to call balls and strikes instead of a human umpire. Even the casual fan know Pitch F/X, even if they don’t recognize the name. That little square in the bottom corner of the TV broadcast depicting the placement of hte pitch in relation to the strike zone? That’s Pitch F/X. If you’ve played around with the the gameday functions within the inimitable MLB At-Bat app, you might have come across data on velocity and trajectory of individual pitches. That’s also Pitch F/X.  As the above-referenced article notes, MLB installed Pitch F/X into every stadium in 2008. I think of Pitch F/X as a bunch of high speed cameras and some software that does physics calculations.

 

I don’t know why, but when I first heard about these Pitch F/X ump games, I immediately assumed that these would be the first professional games ever without a home plate ump. But I was wrong on that point, as there was still a home plate ump present, because obviously, there are still calls to be made besides pitch placement, such as plays at the plate, check swings, batter time outs, etc.  More importantly, however, I knew they would be the first games to run the incredible experiment of having a completely accurate, impartial, and unwavering strike zone for every pitcher and batter. Wonderful baseball podcaster Paul Sullivan has discussed this fantasy on at least one occasion. I’m paraphrasing him when I say that, as instant replay has made the game better, the prospect of eliminating the subjective, biased strike zone would make it better still. And the technology exists to do it.

 

PitchF/X was employed Tuesday and Wednesday, July 28th and 29th, 2015 in two games between the San Rafael Pacifics  and the Vallejo Admirals. The games were fundraising events  spearheaded by former MLB OF (and very cool guy) Eric Byrnes.  (I tend to associate Byrnes with his years crashing into outfield walls with the Oakland A’s and to his post-retirement stint playing softball for the Dutch Goose in Menlo Park, CA. I was at Stanford at the time, so I knew the bar and thought this was incredibly cool. Now, of course, copy cats like Yankee great Hideki Matsui play softball too.) The goal was to raise money for the Pat Tillman Foundation and, as I learned during the game, Byrnes donated hundreds of dollars for each strike out, which added up to a sizable donation. Kudos to him. But I was there for the Pitch F/X experiment as were many other baseball nerds and journalists in the crowd, easily identifiable among the crowd with laptops open wide.

 

It was my first time attending a Pacific Association game, so first a few thoughts about overall experience. It was great. The drive was a  reasonable 20-25 min from San Francisco, though one will likely hit evening commuter traffic on a week night, as I did. The ball park, Albert Park, sits snugly within a larger community park, and it reminded me of a few Single-A parks I’ve been to. It probably sits around 2,500 fans and there is premium on-field seating that allows fans to essentially share space with the  players. There are no proper dugouts; the players sat on plastic lawn shares in the vicinity of the on-deck circles like a classroom of restless pupils. There are no bad seats in the house and I loved the small town, family atmosphere of the game, i.e., the local business advertisements, young folks working the refreshment stands, the local girl who sang the national anthem, and the players being within arm’s reach of fans and kids. I bought a general admission ticket and settled in a few rows up from behind home plate. Personally, I didn’t love the protective netting that spans the entire grand stand because it’s visually disturbing for me to constantly look through it.  That said, I do recognize its value for fan safety.

 

 The quality of play was pleasingly good. I didn’t look them all up, but clearly the players were guys who played college ball or were drafted and played in the minor leagues. I saw hard hit balls, a home run, and some acrobatic plays in the field. A couple notable names for you aficionados: Admiral’s 1B was Lydell Moseby, son of the great OF Lloyd Moseby. The Admiral’s DH was PJ Phillips, brother  to current Reds 2B, Brandon Phillips. Phillips was one of the older players, but he made solid contact and anchored the Admirals order.

 

 OK, what was like to have PitchF/X call balls and strikes? It was surprisingly unobtrusive.  It didn’t disrupt the pace or even the feel of the game. Here’s how it worked: Before each pitch, the umpire crouched behind the catcher as usual. Byrnes sat in the stands with a monitor in front of him with a simulcast showing the pitch placement, just like one sees watching a television broadcast. As the pitch arrives, Byrnes would simply speak the call into a microphone for all to hear – ball or strike. There was the ump, but Byrnes was called the “arbiter”. It was simple, even as implemented in such as technologically-deprived setting as Albert Park. Imagine what a Major League park could do with their Jumbotrons and secondary scoreboards placed around the stadium.

 

I went to the game thinking what a neat experiment I would see, never believing in a million years that MLB would ever adopt such a system. Now, having seen it in action in an understandably crude form, I feel differently. I could easily see a Pitch F/X system working in a MLB game. After an inning or two, I more or less forgot it was there.

 

Here’s an video example of how the system worked:

Early in the game, the home plate ump would instinctively make a pitch call. It was fantastic because, playing to the crowd, Byrnes would say over the PA, “Do you think you were right?……Barely!” A whoop from the crowd. Other times, when a batter struck out looking, or took a strike, Byrnes would make the call and tell the batter exactly where the pitch was. He also did this to the pitcher when he didn’t  get a call; “Sorry, man, that was about three inches outside.” It was fantastic. Not only would this system essentially make strike zone dispute go extinct, it could also be used as an incredible teaching tool. Batters and  pitchers could sharpen there perception of the strike zone, either in real-time during game as in summary form after specific games or many games.

 

The arbiter’s view of the PitchF/X feed.

Want proof? Because Pitch F/X data have made human umpires better.  That’s the main takeaway from the article by Davis and Lopez I mentioned at the start of this post. Since 2008, with feedback on performance using Pitch F/X data, umpires have shown steady improvement in the ability to accurately call balls and strikes. Well, actually, the data show accuracy improvements on strike call, while ball calls have stayed fairly steady. Importantly, while there is considerable variability across umpires in their performance accuracy, all umps appear to show an improvement. That is, the best have gotten better and the worst have better too (at calling strikes accurately).

 

BUT….. even so, the very best umpire, Lance Barksdale, calls pitches accurately at just under 90% of the time. As Noah and Lopez point out,

 

“The difference between Barksdale and a league average ump is about five correct calls per game; the difference between Barksdale and the league’s worst umpire is closer to 10 calls a game. On average, that’s about one judgment call per inning that a good ump is getting right and a bad ump is getting wrong. That might not sound like much, but if once every six outs a batter gets another swing after a third strike that wasn’t or a pitcher strikes a hitter out on a pitch that’s actually a ball, you can start to see the impact.”
What Davis and Lopez are getting at is the benefit of having the best ump versus a league average ump (5 call improvement) or versus the league worst ump (10 call improvement). But how big would the improvement be if we compared the best ump, Barksdale, with Pitch F/X? Well, we’d capture that extra 10% that he is still missing, which if my numbers are current, is about 15 pitches across a 9 inning game. (The average number of pitches thrown per game, which according to this post sits around 300 per game (150  x 2). I haven’t been able to find the breakdown between “called  pitches” vs “swinging pitches”, i.e., either put in play, fouled off, or swung at and missed. But lets say that 50% of pitches require the home plate ump to make a ball-strike determination. That would mean 150 pitches across both teams. Even with Barksdale behind the plate, 10% of pitches, or 15 pitches, would be called inaccurately.) I agree with Davis and Lopez’s argument that the performance gap of 5-10 pitches per game between umps strongly suggests that the best umps, e.g., most accurate, should be given preferred assignments in big games, like playoffs.  But if you believe that, then I hope you will be compelled by the possibility of an even larger improvement in pitch calling accuracy by adopting a Pitch F/X system.

 

But what about the umpire’s union? Surely they’d never go for this, right?  Well, I actually spoke with the home plate ump after the Pacifics game to get his impression on what it was like to be relieved of his pitch calling duties. His answers? A little odd due to force of habit, but it was fun and he could get used to it. There is still lots to do as umpire to earn your paycheck.  Maybe this really could happen, I thought to myself………

 

One other note about the game:
Around the 4th inning, something occurred to me: there were no defensive shifts. It didn’t matter who was batting, the defensive positioning was straight away. How rare in today’s game?! If you follow MLB as closely as I do, you know that defensive shifts have become a regular, accepted part of the game (which is fine with me – it’s evolution baby!) To see teams not shifting now seems odd. But for the four teams of the Pacific Association, how could it be any other way?  Sure, they see each other a lot, but they must rely on “traditional” scouting reports, perhaps limited data, and definitely in-game experience. Not enough to support defensive shifts.
Update on this old post: 
MIT Technology Review recently published a brief article entitled, Baseball players want robots to be their umps. The article notes that Chicago Cubs super-utilityman Ben Zobrist is a “vocal supporter” of automated balls and strikes calls.

“That’s why we want an electronic strike zone.”

Zobrist, shortly before getting his first career ejection

This is the first time I’ve heard of an MLB player explicitly stating a desire to move away from human umpires on this issues. Now three years since I wrote my initial post, I’d say that it’s inevitable that human umpires will not call balls and strikes in the not-so-distant future.  Instant replay in baseball is nearly 10 years old now, having been expanded to cover more of the game’s trickier situations (close calls at bases, fair/foul, etc). Frankly, it’s made the game better. Self-driving cars are the natural culmination of the progression of automation in car driving that’s been going on for decades. I predict that “robot umps” will be the culmination of technological progression in baseball, with the fringe edge cases still requiring human review.

Fenway franks

 

I’m sitting in a Starbucks near the Boston Back Bay train station.  It’s a Monday morning, and I have an hour or so before my train to New York City.

I’ve been in Boston for nearly four days, the trip centered around a set of baseball tickets I “won” in an auction. I italicize the word “won” because no one else bid on the item, which is a bit like taking the gold medal when there are no other competitors. The ticket package was a set of tickets for two to three games between the Yankees and Red Sox at Fenway Park. I’ve long hoped to visit Fenway Park, and I’m turning 40 this year, so I thought I would treat myself with this trip. My sweetheart joined me for a good part of it, and it was a great weekend to be a Yankees fan, as they swept the Red Sox over the three games; 3-2 on Friday, 4-2 on Saturday, and 8-5 on Sunday.

Thoughts on Fenway Park

The Stadium – For some reason, the park didn’t feel as old as it is, in the way that Wrigley Field felt old (and wonderful) when I first visited it last year. I’m not ashamed to admit that when I first emerged from the tunnel at Wrigley and took in my first view of the field, the flags, the famous scoreboard and ivy, I had genuine tears in my eyes. I don’t know why it happened – I was excited and giddy when I approached the stadium, not weepy or emotional. I have no special connection to Chicago, the Cubs or Wrigley Field. In retrospect, the best explanation I have is that gazing out from the stands at Wrigley made me feels as if I stepped inside one of my favorite movies, Robert Redford’s The Natural. Whatever the reason, I definitely sensed that the stadium was old, a classic, a landmark, a shrine.

I didn’t feel the same way inside Fenway Park.

Walking the stadium, I looked closely at the brick facade, and over the course of three straight games, I read as many plaques and momentos as I could. Settling into our seats (on a cold, cold Friday night – sorry, babe!), the stadium felt nicely-sized, but not elderly. Present are all the modern accoutrements – three large Jumbotrons with tastefully-styled projects to match the look and feel of the scoreboard on the Green Monster. But the Green Monster itself, to me at least, did not seem old; it looked freshly painted, adorned with a new set of logos for this year’s advertisers. In my mind, I envisioned a wall with the wear and tear of decades of baseball, with innumerable dents and scars. If those blemishes still exist, they have been covered up with considerable make-up.

We sat close to the field, halfway between Pesky’s Pole and the visitor’s bullpen. I would not have purchased these seats myself, but we did not know the ticket locations prior to bidding on them. (Short digression: Here’s another lesson in the perils of making assumptions. I assumed the seat would be these incredible box seats, perhaps even catered (!) because the auction listing claimed the 6 tickets had a total value of $1,200. I bid $500, which was the starting bid price. At that point, I’m thinking to myself, “I’m going to get some incredible tickets at a great price!” In fact, the tickets had a face value of $70 each; I overpaid. This is Yankees – Red Sox 2015, not 2003, so one cannot argue that the face value is way below market rate.)  The vantage point from the seats took some getting used to, as we were essentially straight down the first base foul line. My sweetheart had trouble with this. She repeatedly missed the action at home plate because she thought the on-deck batter was the actual batter. Even I had trouble gauging the trajectory of batted balls, and making a fair/foul call down the left field like was impossible.

But over the course of three games, I grew accustomed to the vantage point, and began instinctively turning my head to the main Jumbotron scoreboards to catch the replay (which is shown after every play, which is great) or glancing to the auxiliary scoreboard along the  grandstand or the classic red and green lights at the base of the Green Monster to check the count. I like being close to the Yankees bullpen, our sight line being directly behind the pitcher to the catcher. I believe it was the first time my sweetheart had ever seen the break of curveball.

A few things I found odd or disturbing about  Fenway:

  • The odd: I never saw where, on which scoreboard, the official scorer’s play call was displayed. For example, when a pitch got past the catcher, I never saw where the official scorer’s decision of passed ball or wild pitch was displayed. The same went for errors and stolen bases. At other MLB parks I’ve visited, the scorer’s call is prominently displayed on the main scoreboard soon after the play in question. That definitely wasn’t the case at Fenway.
  • The disturbing: The ushers do very little policing of fan seating, seat switching, or unruly behavior. My experience at Fenway was, by and large, absolutely fine in terms of fan experience. Perhaps it was my particular section, which was in the outfield and not the luxury seats. That said, I saw undesirable fan behaviors at Fenway that I’ve never seen at other parks, and that includes Oakland Coliseum, old and new Yankees Stadium, and many others. First, yes, the usher gladly helped you find your seat, and if other fans were seated there, they would move. But I’ve never seen such relentless seat switching. Fans were constantly dropping into vacant seats near us, watching an inning or two, and then leaving. Second, the ushers did zero traffic control of fan movements to reduce disturbances. At AT&T Park, for example, the ushers will hold back fans returning to their seats until a break in the action occurs, generally after an at-bat. Fenway Park ushers do not so this, so there were innumerable stand ups/sit downs and blocked views while watching the game.  Lastly, I was personally heckled a few times during the Sunday night game. It wasn’t a big deal, and I realize that every section of the stands has its own cast of characters, its minor soap operas.  But in the nearly 100 MLB games that I’ve attended across 14 ball parks, I’ve never been heckled by nearby fans for keeping score on a scorecard, which I generally do. (“You know, the ESPN app will do that automatically for you?” “What an idiot! The guy is keeping score on paper.” “Get in the 21st century!” “Hey, did you get that?” “Hey, how did they make the first two outs this inning?”) Yes, it was a couple of young guys drinking beer, and no, it didn’t really bother me because I simply ignored them and that’s no fun. But they did it a couple of times, and ultimately I felt a bit uncomfortable turning my head to look at one of the scoreboards, lest I draw the attention of one of the hecklers seating in that direction. So, following the lead of so many other Fenway fans, I simply got up and moved to much better seats in the 7th inning.
  • Singing Sweet Caroline during the 8th inning is peculiar and has seemingly nothing to do with baseball or the circumstances of the game. But I’m not so much of a grouch that I didn’t eventually come to enjoy it. The words are uplifting; the “so good, so good, so good” chant. It’s no more unusual than the SF Giants playing Lights by Journey every night. If you know me, you know that one thing I absolutely love about the game of baseball is the tradition of singing Take Me Out to the Ballgame during the 7th inning stretch. And I loved the chanting and singing I witnessed at the professional ball games I went to in Asia, specifically a Doosan Bears game in Seoul, Korea and a Yomiuri Giants game in Tokyo, Japan. So, if I liked those traditions, I should like the adoption of Barry Manilow at Fenway.

This was one of my favorite trips to Boston. My sweetheart and I sampled many coffee shops around Harvard, MIT, and BU. (I love being around students. It reminds me of my graduate student and post-doctoral days.  I easily slip into reveries with big projects and future plans taking shape.) We strolled down Newbury Street. We biked around using the Hubway bike share, which was incredibly efficient and fun, taking us from Back Bay to Somerville and everywhere in between. We went jogging along the Charles River. We made love in the afternoon. We reflected on our beautiful son.

Thanks for a memorable weekend, Boston…..