A view from the front lines of baseball's data analytics revolution

A view from the front lines of baseball’s data-analytics revolution

In the second of a two-part interview, Houston Astros general manager Jeff Luhnow describes how analytics has changed baseball—from Moneyball in 2003 to the last out of the 2017 World Series.

“Whenever people talk baseball,” observed beloved Hall of Famer Willie Stargell, they don’t say, ‘Work ball.’ They say, ‘Play ball.’”

Perhaps. But for people who perform at baseball’s highest levels—the athletes, coaches, scouts, and executives who compete for a living—the stakes are considerably higher. In fact, a major-league umpire doesn’t even say, “Play ball.” Instead, he’ll typically point to the pitcher’s mound at starting time and announce, abruptly, “Let’s go.” The game is a business. And the business is to win.

Video
Baseball’s next biggest wave

Fifteen years ago, Michael Lewis, in his book Moneyball: The Art of Winning an Unfair Game (W. W. Norton & Company, 2003), described the growing importance of data in the baseball business. That same year, Jeff Luhnow joined the St. Louis Cardinals as the team was starting to embrace the ideas behind Moneyball. In this interview, Luhnow reflects on the course of baseball analytics over the years, and the contributions of the Astros along the way. This is the second interview of a two-part series. The first is for analytics leaders seeking inspiration in the Astros’s transformation. This one is for baseball fans, whose ability to grasp the modern game depends on an understanding of analytics’ growing role.

The Quarterly: Let’s go back to 2003, when you entered baseball with the Cardinals. What was the state of play at that time with respect to data analytics versus traditional scouting?

Jeff Luhnow: Our sport has been consistently the same for decades and, quite frankly, a century. There have been some changes in terms of mound height and some of the rules around the game, but it’s essentially the same. What’s really changed dramatically in the last 30 years is information, and how that information is used to make decisions. This started in the ’50s, when some innovative baseball people started to recognize that the traditional ways of evaluating player performance were not accurate. Branch Rickey wrote an article in Life magazine about the early version of on-base percentage. That continued to evolve and really started to accelerate in the ’90s and going into the turn of the century. Certain teams recognized that there was value in the information and in how they could use the information.

There’s no better example of that than the Oakland Athletics, who were a team that was struggling revenue-wise compared with the big-market teams. They were able to utilize the information to evaluate players in a different way, to capture an edge versus the bigger-market teams. They became one of the more successful teams around the turn of the century and made a lot of playoff appearances, in large part because they were finding undervalued players that they were able to recognize through their use of information and analytics. That was all captured in Michael Lewis’s Moneyball, published in 2003.

It’s interesting. In our industry, you could write a book about someone whose best practices are allowing them to compete in a unique and advantageous way, and yet most of the industry would pretty much ignore that and keep doing things the way they were. One person that didn’t ignore that was Bill DeWitt Jr., the principal owner of the St. Louis Cardinals since 1996. When he saw Michael Lewis’s book, he thought this was an opportunity for the Cardinals, who were a larger-revenue, larger-market team, to capture an edge. In 2003, Bill DeWitt Jr. decided to bring me in to the Cardinals and begin the exploration of how to transform into a modern organization that utilizes information.

And over time, there have been teams like the Boston Red Sox, the St. Louis Cardinals, and the recent Chicago Cubs and Houston Astros that have used information and analytics as one of their competitive weapons and have been successful in not only reaching the playoffs but also actually achieving championship status. That was one of the criticisms about Oakland. They could get to the playoffs, but the analytic approach didn’t help them ever win in the playoffs. But that changed when the Red Sox and the Cardinals started to win the World Series and the industry knew that they had a strong analytic bent. And obviously, in recent years, it’s been even more so with the Cubs and the Astros. Teams have always looked to those teams that succeed, that win championships. What did they do that we can do in the future?

The Quarterly: How would you contrast the information in 2003 with what’s available now?

The image above and the ones that follow in this article were captured from video shot at the Astros’s spring-training facility, in West Palm Beach, Florida, on the day of the Quarterly’s interview with Jeff Luhnow.

Jeff Luhnow: Oh, it’s worlds different. Back then, when I talk about analytics and information, it was looking at historical performance to predict future performance—very much like in the financial-services industry, where you’re looking at the performance of stocks and the economy and trying to figure out what factors are predictive and sticky and what the future might look like. At that point, there was very rudimentary baseball information, like walks and singles and doubles and strikes and balls. But there was a lot of predictive value inside of that, and you could begin to build models that would help you predict a certain player at a certain age that plays a certain position that has had this type of résumé in the minor leagues and in his early major-league career. This was roughly what we would project him to do next year and the year after.

Those projections became much more accurate than if you just used traditional methods of asking your scouts or your baseball evaluators: How many home runs is this guy likely to hit next year? How many runs is he likely to score? But while those projections ended up being substantially more accurate and predictive, they still have a huge error bar around them, because we’re trying to predict what human beings are going to do on a field of competition, and there are so many other variables that you can’t control for. That was the breakthrough in 2003 around that time.

Today, it’s completely different. We now have so much technology around the ballpark and information about the trajectory of the ball, the physics of the bat swing, the physics and the biomechanics of the pitcher’s delivery—so many components now that advanced sciences have worked into our game. It’s, quite frankly, overwhelming in terms of the amount of information that we have access to and intimidating to figure out how to analyze all that information, work through it, and come up with the takeaways that will allow you to continue to do what we tried to do back in 2003, which is to make better predictions about what players are going to do in the future on the field.

The Quarterly: What other changes jump out at you over past 10 to 15 years?

Houston Astros

Jeff Luhnow: In 2003, there were maybe four to five clubs that had analytics-dedicated people on their payroll, and typically they were in an office down the hall working on recommendations to people who may or may not pay any attention. I think what’s changed today is that every general manager has some background or interest in analytics, and the typical size of the group in the front office is probably somewhere between 12 and 15 full-time people who all have advanced degrees, whether it’s computer science or physics or mathematics or some other discipline. Along with that, there are data departments in organizations. Most organizations now have database folks and data scientists that are on their payroll and that are helping them not only store the information and organize it properly but also evaluate what it means.

There was also a trend in the past of using external companies to house data, like scouting reports or statistics. Most of that has now come in-house. When I was with the Cardinals, we used an outside provider, and when I got to the Astros, they were using an outside provider, but the response time and the customization was lacking. Most important, when you come up with a way of looking at the world and you want the external provider to build the model for you, you don’t want them to share it with the other 29 clubs. It’s difficult to have the confidence that it’s not going to be shared in some way, shape, or form. I think that’s led to most clubs believing that their way of handling data and information is a competitive advantage. It therefore becomes critical to have control over that in-house.

The Quarterly: What were your immediate priorities when you joined Houston in 2011?

Jeff Luhnow: When you’re lagging, the first thing you need to do is figure out how to make sure you’re not losing ground anymore. In 2011, things were changing pretty rapidly in terms of the types of information that was out there, the types of analysts that you might want to hire, the data scientists—all of that. We felt we had an advantage relative to the infrastructure that had been built in St. Louis over the prior eight years. We had a clean slate.

So we were able to start with a fresh piece of paper and say, “OK, given what we think is going to happen in the industry for the next five years, how would we set up a department?” That’s where we started, “OK, are we going to call it ‘analytics’ or are we going to call it something else?”

We decided to name it “decision sciences.” Because really what it was about for us is how we are going to capture the information and develop models that are going to help the decision makers, whether it’s the general manager, the farm director who runs the minor-league system, or the scouting director who makes the draft decisions on draft day. How are we going to provide them with the information that they need and that will allow them to do a better job?

The Quarterly: Did you have to hire new people?

Jeff Luhnow: Yes. I was fortunate; the Cardinals allowed me to bring one of the people I had worked with in St. Louis, so I had somebody that had done this before and was eager to do it again with a clean sheet of paper. I think we hired three people—a database person, a data scientist, and an analyst—within the first couple months and just started cranking away. The group has now grown to about 15 people. But that first year, it was three or four to get it done.

The Quarterly: We hear a lot about fail fast, learn fast. Were there any failed experiments that stand out for you?

Houston Astros

Jeff Luhnow: We’ve made plenty along the way, and you have to own those mistakes. We made mistakes at the top of the draft. We made mistakes getting rid of players that we shouldn’t have gotten rid of. There was a player in our minor-league system who got into the big leagues, J. D. Martinez. Great guy, loved him, great player in a lot of ways. He played three seasons for us in Houston, from 2011 to 2013, and he just wasn’t developing into a regular everyday player. That winter, he made some improvements on his swing. He came and talked to us about it, but we really didn’t give him an opportunity to show us what he had done in the spring of 2014, so he didn’t get enough at-bats. Then it became a roster crunch at the end of spring, and rather than just send him back to the minor leagues and have him be frustrated there, we ended up cutting him loose, just releasing him so he could catch on with another team, the Detroit Tigers. He went on a tear there and became an all-star player, and he just signed with the Boston Red Sox for $110 million. We let him go for free. But I did learn a lot: There are meaningful changes that players can make during the off-season that can change their profile going forward. And that’s a lesson learned for me.

And fortunately, we had a high hit rate on a lot of the decisions that we did make relative to other clubs, and we were able to improve our system very rapidly. You have to realize we’re playing probabilities. There are going to be some losses along the way. There are going to be some decisions that don’t work out. And you have to continue to emphasize that we’re playing a long game here. We’re making multiple bets. Because one bet doesn’t work, that anecdote is an anecdote. It’s just that, and you can use it.

The Quarterly: When did you start to feel, “Yes, the system is working”?

Jeff Luhnow: In 2013, the year we lost 111 games, that was the most losses that the Astros had ever had in a year. It was absolutely brutal. I went to every single game. I wore every single loss. I took it personally.

But the good thing about what was happening in our system was that we knew the underlying metrics were all improving at a very impressive rate. Our farm system was developing quickly. Our institutionalization of analytics was taking root. Our minor leagues were winning with good players that were young for every level. We knew that it was just a matter of time before that worked its way into the system and got to the big leagues. We were also training our coaches in the big leagues to be receptive of the new type of player, the new type of information.

For me, the key was 2014. We called up George Springer, and we ended up improving our win–loss record that year by 19 games. There were enough periods during the year where we were taking on some pretty strong teams, and winning that, it was clear that were good signs for the following year. And lo and behold, we get off to a fast start in 2015, and we end up getting to the playoffs that year.

The Quarterly: Last year was obviously pretty magical. Can you recall some moments when you looked out and said, “Wow, analytics did that?”

Houston Astros

There are so many elements of our strategy from 2011 to 2016 that manifested themselves on the field in 2017. Alex Bregman was not the consensus top pick in the draft. But we had him as the top player by a healthy margin. We were fortunate that he wasn’t selected first; he fell to second. We got him. He was a huge part of last year. Carlos Correa was a consensus 10 to 15 pick in the draft in 2012. We had him as the number-one player by a pretty wide margin. We selected him first overall that year. Surprised some folks. He ended up being one of our better players, Rookie of the Year, and a big part of our championship season. Dallas Keuchel: I mandated to the staff that he was going to have to make the team by 2014. He ended up having a breakout year and winning the Cy Young the year after that.

In terms of our style of play, we continue to be the team that is most on the outer bounds of shifting [our infielders into nontraditional positions on the field; for more on shifting, see “How the Houston Astros are winning through advanced analytics”]. Our infielders were in a position last year, after having now done it for three or four years, to turn a lot of double plays. We had a terrific defensive output. We captured a lot of extra ground balls. That, combined with the fact that we look for ground-ball pitchers like Charlie Morton and others that we’ve brought in, gave us extra runs saved that we hadn’t had in years before.

Also, we have a program in the minor leagues, which is based both on health and on performance, where we ask our starting pitchers to piggyback. Instead of one starter for each game, we have two starters each game. That allowed them to get experience late in the game, allowed some pitchers to finish games, and allowed pitchers to get used to coming into games in the fifth inning. If you look at the American League Championship Series game seven, Lance McCullers was one of these pitchers that we developed in our system in a tandem rotation, and he came in during the fifth or sixth inning and closed out the game. He had done it for us a few times in the minor leagues, so it was kind of neat to see that happen.

I think the best example is the final play of the 2017 World Series. We’ve got a catcher who is good at framing—where you can turn a ball into a strike by how you present the pitch to an umpire. We’ve got a pitcher on the mound, Charlie Morton, who was a starter finishing the game; we had identified him in the free-agent market because he had some stuff that we thought could play up in our system. We had an infielder who was standing in right field, a position where five years ago no infielder would ever stand.

Of course [Dodgers player Corey] Seager [the last batter of the game] hits the hard ground ball right to Jose Altuve. Altuve is standing in right field, fields the ball in the shift, throws it to Yuli Gurriel, a guy we found in our international scouting out of Cuba who took a lower-dollar offer to be with the Astros and could’ve gone to another team. There was a lot in that final play. We were in the shift. We used tandem pitching. The pitcher was a guy we had used analytics to identify, and we had two players that had accepted less money to come to the Astros. And that last play of the World Series, which led to our first championship ever, encapsulates all of that.

About the author(s)

Jeff Luhnow is the general manager of the Houston Astros. This interview was conducted by Aaron De Smet, a senior partner in McKinsey’s Houston office, and Jeff Hart, a partner in the Houston office.

Related Articles