Small Ball Meets Big Data: Inside the Twins’ Swing Toward Analytics-Driven Decision Making
On the green fields of Florida, on a temperate winter day, the 150-year-old game of baseball’s reemergence implies permanence and tradition. Kids and adults line up for autographs from star players while minor leaguers, some spending their first year in the U.S., nervously navigate warmups.
But behind the faÃ§ade of the familiar, a rethinking is underway, replacing values like “clutch” and “a gamer” with a metrics-based approach. It is an evolution that mirrors the one taking place in the business world, where technology has brought with it reams of data and changed the understanding of how companies win.
So this winter the Twins are preparing for the season somewhat differently. Batters are practicing an uppercut swing at a hitting cage designed to reject ground balls. (New designated hitter Logan Morrison was signed out of free agency after a 2017 renaissance built on the uppercut.) Outfielders are practicing route efficiency to reach hit balls faster and cut off extra base hits. “We’ve redesigned our drills for that phase of play,” explains Twins coach Jeff Pickler.
These ideas are rooted in “Statcast” data that Major League Baseball has been collecting at stadia since 2015. Pickler is weary of the “revolution” talk, seeing it as just another evolution of the game. “It works best when baseball people ask questions that lead to research that leads to players,” he says late one afternoon in the Hammond Stadium locker room.
Old Days, Old Ways
Since the glory of the 1991 World Series, where a team of scrappy “gamers” turned clutch and grit into legend, the Minnesota Twins have finished 10 games above .500 only four times, with zero returns to the fall classic. Even the so-called glory seasons of the ’00s, with six division titles, generated a humiliating 8-22 playoff record, where the Twins showed themselves to be hopelessly outmatched against the game’s best teams.
Cue the move to Target Field in 2010, marketed to taxpayers to allow the Twins to create a revenue base for permanent competitive balance. A decade in, their playoff record is 0-4, with one division title and four last-place finishes, plus the notorious 59-103 2016 season, the worst in team history.
That July, longtime general manager Terry Ryan—a favorite within the organization and MLB, known for his gentlemanly old-school approach—was issued his walking papers, as the team went in search of new leadership and new approaches.
In less tradition-bound baseball organizations, a revolution had already taken place, rooted in the ideas of the Oakland A’s under general manager Billy Beane. Moneyball, written by financial journalist Michael Lewis, details how the underfunded yet overproducing A’s used data to identify inefficiencies that it could exploit to even out its economic disadvantages.
Moneyball, later a hit movie starring Brad Pitt, is widely misunderstood among the public and even within the game as the story of baseball’s subjugation to computers. Yet it is a reasonable touchstone for the early years of an era of rethinking the game that found the Twins, in 2016, among a small handful of teams on the outside looking in.
The Twins were a scouting-dominant organization—the observational skill of sage veterans was valued above all and guided player personnel decisions. But those decisions were failing the team, as organizations with less revenue and poorer facilities (Tampa Bay, Oakland) continually out-drafted, out-traded and out-played them.
“Baseball was late to understand the role of cognitive bias,” explains Ryan’s successor, Twins chief baseball officer Derek Falvey, referencing the truism “the eyes don’t lie,” and the sense that teams were making poor decisions by relying on human factors with built-in flaws.
Falvey arrived to implement a system that is less about a single method or approach, but a discipline built on discipline. “We’re trying to root all our decisions now in evidence. Can we subject our opinions to a test to verify them?” he asks. “Can you create a system to help people make decisions and scale it to the entire organization?”
Analytics Extrapolated: The Uppercut Swing
For a long time now, the mantra in pitching was that effective pitchers threw “down in the zone”— sliders, sinkers and cut fastballs. The theory was that a low pitch was hard to drive out of the infield. “The down-in-the-zone pitch had been accepted as the gold standard,” explains Twins director of baseball operations Daniel Adler. But things were about to change.
The Astros and Dodgers used MLB Statcast data that indicated low pitches hit golf-style had an enhanced propensity to be home runs. Suddenly, the pitches most batters avoided became desirable to teams that had crunched these numbers.
“The uppercut swing is like the three-point shot in the NBA,” explains Adler. “When they go in, you get 50 percent more points. Fly-ball-based hits become home runs, which justify [a swing that gives up the opportunity for] ground-ball-based hits.”
Read more from this issue
But the Astros and Dodgers excelled not just because they alone had the data.
“The uppercut swing phenomenon required buy-in,” explains Adler. “It required coaches who trusted the data [and] who had good relationships with players. Asking spectacular athletes to change what [they] do and how [they] do it is not easy.” (He notes that newly signed DH Logan Morrison remade his swing last season with dramatic improvements in power as a result.)
And it requires skill, as well. “If you can’t [hit] the ball out [of the ballpark], it’s just a fly-out swing,” notes Twins’ bench coach Jeff Pickler.
Post-World Series, everyone knows about the uppercut; but not every team can or will adapt. And smart pitchers will certainly counterattack. “You can only optimize for the present,” Andres says. “The next evolution is pitching up in the zone” to combat it.
But baseball is a cat-and-mouse game, and “data won’t ever change that,” says Falvey, “so you are really creating a learning organization rooted in a philosophy.”
The Twins’ inability to develop pitching, the foundation of the game, made Ryan vulnerable. Falvey was hired from the Cleveland Indians at age 33, in part for his reputation as a pitching savant, but also for his holistic take on how the organization had to change. “Derek and Thad [Levine, the Twins’ new general manager] have worked hard to break down walls within baseball operations,” says Twins president Dave St. Peter, “ensuring that scouts and player development staffs are working hand-in-hand, sharing ideas, etc.”
The Twins are quick to point out they were not running a 19th-century operation that stood willfully in opposition to the game’s evolution. The Twins did have data analysts, notes Boston University professor Andy Andres, who has long studied analytics and sports, “but it’s not clear how much it impacted decision-making [under Ryan]. It’s pretty clear they’re now serious about it.”
The Twins were hamstrung not just by an old-school, siloed approach, but also a lack of front-office talent. “Dating back to the process to find new leadership of our baseball operations, every candidate we interviewed shared a vision to grow our internal team,” St. Peter says. “From a pure staffing standpoint, the Twins baseball operations was smaller than almost any team in the game.”
The organization did not approach its effort to change with characteristic frugality. “We’ve invested millions of dollars in incremental staff, systems and technology—all aimed at ensuring we are in the best possible position to build a perennial championship contender,” says St. Peter.
Though the philosophical tug-of-war over baseball’s soul (metrics vs. observation) is old news, “sports was not on the leading edge of big data,” insists Dan Atkins, executive director of MinneAnalytics, a nonprofit serving the state’s data science and analytics community.
Analytics first came to the fore in sports among a community of baseball fans looking for new ways to think about the game’s foundational statistics, many of which offered a distorted view of why some players were effective and some weren’t. “Some of it was taking existing statistics and coming up with novel insights,” says Andres.
An early breakthrough was a way to define a baseline player’s performance to measure other players against. “[Data journalism pioneer] Nate Silver’s ‘value over replacement’ was an early effort to judge who is worth signing,” says Atkins.
WAR (Wins Above Replacement)
This all-inclusive stat has become a darling of the baseball metrics community. It attempts to define how many additional wins a player is worth over the course of a season compared with a “replacement-level” player—a player who would not command a premium above the major league salary minimum.
Next came WAR, or wins above replacement, which attempts to value player contributions in terms of the additional wins they represent per season.
Falvey’s arrival coincided with the broad-based emergence of Statcast, MLB’s data measurement tool. Atkins says in Statcast’s first game, at Wrigley Field, “the Cubs collected more data than they had in a century plus of baseball history.”
This is where the “big” in big data comes in, and for the first time in its history, Major League Baseball has more data than it can manage. Which explains why the Twins say one of their biggest expertise gaps is in computer programming. “They need it,” says Atkins, “to reduce the data to something usable.”
Remember when Target started offering coupons for pregnancy-related goods to women who weren’t pregnant, or at least thought they weren’t? “Target studied the buying history of women who were of childbearing age. They found a ‘tell’ in the pregnant women versus a control group,” explains Atkins. It allowed the software to, in some cases, make an educated guess that a woman was pregnant even before she knew.
Atkins says this mirrors the contemporary search for relevant data in sports. “So you can’t predict how Joe Mauer will do versus a pitcher he’s never faced, but you can input a bunch of factors, including how Joe fares against like pitchers and come up with a pretty good guess.”
The Twins are racing to bring their analytics up to league average while finding weaknesses or “tells” to exploit before other teams discover them. Asking Falvey and his team to discuss these modes and methods elicits mostly vague values and themes.
FIP (Fielding Independent Pitching)
FIP is an attempt to improve on earned run average in evaluating pitchers. All pitched outcomes that involve hit balls (other than home runs) are affected by the quality of a team’s fielding. Pitchers who play in front of better fielders have lower earned run averages, but may not be better pitchers. FIP tries to make that clear.
“The rise of analytics has made teams more secretive,” says Andres. They won’t tell you what they’re looking for in a player because “they believe they possess something proprietary.”
“In terms of exploiting inefficiencies, one of the things I struggle with in baseball is where are the competitive advantages to be found?” says Daniel Adler, the team’s new director of baseball operations.
Take the technique of “framing,” where the catcher positions his glove in a way that elicits a more favorable strike zone. “Pitch framing started as something observed by scouts, but it was pooh-poohed by analysts because of the difficulty in measuring it,” says Adler. “We tend to be dismissive of things we can’t measure.
“The scouts were right, and it came into broad acceptance, but there were only a few years to make hay,” Adler notes.
“Huge market inefficiencies close quickly now,” says Andres. Teams discover “ ‘Holy crap, this is real.’ It gets public. and all teams figure it out. The Twins were late to pitch framing because they didn’t have the scouting or analytics to employ it.” Or they doubted it.
The Twins were expected to be more active in this off-season’s constipated market for free agents. “There were free agents we thought would be inexpensive because we thought we had identified aspects of their game that other teams hadn’t,” says Adler, “but they weren’t inexpensive because we weren’t as unique as I thought. The game is evolving so fast that we need to be an organization that can adapt quickly.”
In the past, player agents or other teams could rely on non-analytically minded teams to distort or disrupt the market by filling their rosters with catchers who could not frame pitches or overpaying for free agents whose metrics indicated their game was flawed.
Atkins offers this analogy: “I could count cards in a poker game, but because my brother-in-law is emotional, he bets in ways that force a rational player to act irrationally.”
Falvey sees that era waning. “Teams are creating systems to evaluate free agents. Creating models to try to extrapolate their asset value. Smaller payroll teams are less disadvantaged in an evidence-based environment.”
Twins outfielders Eddie Rosario, Byron Buxton and Max Kepler. (Photo by Brace Hemmelgarn)
Translating the Nerds
As MLB coalesces around an analytics mindset, the Twins brain trust believes a difference-maker is going to be the ability of the nerds to communicate with the jocks.
“Finding things yet to be identified is important,” says Adler, “but being able to utilize the things you already know is more important.”
Put another way, when, in the old baseball paradigm, everyone was a player or former player, the mindset was similar across the organization. Not so much today. “There are two different skill sets at work,” explains Andres. “Playing baseball takes extraordinary skill. Analytics is also an extraordinary skill. Connecting the two is not as simple as you’d think.”
Atkins agrees: It’s not as easy as “the guy from the office says to move 15 feet to the left.” Teams need “translators,” on- or off-field staff who can explain how analytics translate to the game in a way that’s clear and motivating.
“We focus on what is actionable,” says Falvey.
“But we’re not going to players talking about reaction time and route efficiency,” says Pickler. “We talk about how the best outfielders [play] and our drills reinforce that.”
This expertise is not unique to baseball or sports. “Translation is a job description that has emerged in business,” says Atkins. “It involves reading and reaching the decision maker. The translator is rarely the advanced analytics guy who can’t look you in the eye.”
A Career in
Back in the day, a career in baseball usually flowed from playing or coaching the game. But more and more, an on-field background doesn’t much lay the groundwork for a career in the front office.
More and more of the game’s operating leadership has advanced degrees from elite universities (Twins director of baseball operations Daniel Adler holds law and business degrees from Harvard) and little if any time on the field. They come with a focus on developing systems and processes to optimize the game. And they are not grizzled veterans. Terry Ryan is 64; his successor, Derek Falvey is 35 and is far from the only leader in MLB in his 30s.
“There is definitely a shift toward youth because of analytics,” says Boston University professor of mathematics Andy Andres. “The skill sets in play are very different now.”
“The jobs are extremely taxing, and they age people in dog years,” adds Adler.
An MIT degree is not yet essential, however, despite rumors to the contrary. “You don’t have to be great at math, you need to understand what the math is telling you,” says Dan Atkins, executive director of MinneAnalytics. He notes that the University of Minnesota’s Carlson School of Management wants data literacy to become integral at the undergraduate level and that “data scientist” leads lists of the most promising jobs in America.
Falvey worked to make an attractive case to Adler as he finished his degrees at Harvard, telling TCB that executive “talent wars are real” within pro sports. Falvey says the Twins have restructured their baseball operations “to allow new leadership opportunities” to better retain talent.
Adler, who majored in economics and physics, describes himself as intensely “interested in how people make decisions.” He says he chose the Twins in part because he “had been told organizations with new leadership were good places to go.” He was intrigued at the thought of getting in on the ground floor of what Falvey was building in Minneapolis. He leads the Twins’ R&D group and consults on player salary arbitration.
“Daniel is a bright young man,” notes Andres. “Twins fans should feel very fortunate.”
No team has moved more heavily away from human factors as baseball’s 2017 World Series champion Houston Astros, who completely rejected the value of observational expertise last year when they fired most of their pro scouts.
But not everything in the game is optimizable with data, and the Twins don’t plan to follow suit.
“The best scouts judge character,” says Adler. “A guy can have amazing stats, but if he’s a horrible teammate, his ceiling will be limited. Our goal is to be able to quantify how accurate our assessments are, with a goal of weighting our subjective data.”
“These people are still very valuable,” says Andres. “Nobody relies on them exclusively anymore. Data on baseball mechanics is [now] dominated by analytics; data on makeup and flexibility is an observational skill.”
Adler says the Twins’ perspective is that “there’s no evidence the game is good at understanding psychological factors and extrapolating them into outcomes, nor [is it able to measure] resilience and durability. Trying to project [the trajectory of] young players is not data strong.” He adds that the defensive side of the game remains an area where metrics’ predictive ability has not been as reliable, though Adler expects Statcast data to eventually improve that.
The short-term future of analytics, he says, is less about crunching the same data differently, but using the league’s new technological infrastructure to develop “measurables,” he says. “Cameras are tracking everything.”
Human factors will inevitably intervene. “We’re still subject to so many biases,” Adler continues. “Take how we present trades to another team. Should we present data, how much, and when in the process?”
It should be no surprise that analytics may be affecting the game in unanticipated ways.
“Games are probably getting longer because of analytics,” says Andres. As teams fixate on waiting for the right pitch to achieve the correct launch angle and exit velocity, at-bats have become longer, more pitches are thrown and more substitutions take place. A three-hour game is now shorter than average. That’s not necessarily a good thing. “The appeal of long games is waning,” Andres says. “They have a lot of business data on the game and how people watch it.” MLB and its players are now at odds over game lengths.
So at some point the best interests of the game may conflict with the best interest of a team or a player. “We have a greater obligation to the sport and the game than what happens in a specific game,” says Falvey. “We have to be vigilant to unintended consequences.”
Adam Platt is TCB’s executive editor.
Note: This feature appears in the April 2018 edition of Twin Cities Business.