Inside baseball: CMU-Q dean reflects on the past, present and future of operations research
This past summer, Michael Trick, dean of Carnegie Mellon University in Qatar (CMU-Q), a Qatar Foundation partner university, was invited by Amazon to share his insights into the field of operations research.
“They asked me to talk to them about the story of operations research,” said Trick, who was a Carnegie Mellon researcher and educator in the field for decades before coming to Qatar. “And they wanted to hear about baseball.”
As an operations research expert, Trick has consulted the United States Postal Service on supply chain design, college basketball conferences on scheduling issues, and telecommunications organizations on bandwidth allocation. But it is his work with Major League Baseball (MLB)—and how it relates to technology, data, and everyday life—that invites the most questions.
“When you hear the question, ‘what’s the best way to do something,’ that’s an operations research question,” says Trick, “and the story of baseball scheduling is the story of the field.”
How hard can it be to schedule baseball games?
Operations research is a field that bridges computer science with a myriad of different industries. Any area with complex problem sets—shipping routes, military resources, hospital management—likely uses operations research to optimize the way to get the job done. This includes sport scheduling.
In 1994, Trick was a faculty member at Carnegie Mellon when he received a phone call from Doug Bureman, the executive vice president of the Pittsburgh Pirates team. “He asked me if I was interested in looking at scheduling for Major League Baseball. I thought, how hard can this be?”
The problem turned out to be more complex than Trick imagined. There are 30 teams in MLB, and each team plays 162 times, for a grand total of 2,430 regular season games each year. This is nearly 10 times the number of games in the U.S. National Football League, and six times the games in Premiership Football.
Schedulers face many constraints, like each team must play at home half the time, and half of the home games must be on weekends. Teams also submit lengthy requirements and requests for their schedules.
“The constraints mean that a schedule you are building may turn out to be impossible to finish. It takes time to get far enough into a schedule to know if it is even possible,” said Trick.
In 1994, the MLB schedule was prepared by Henry and Holly Stephenson out of their home in Martha’s Vineyard, Massachusetts. The official schedulers since 1981, the couple used a small home computer for some of the basic work, and they balanced the many variables themselves to finish the schedule.
“It took us 10 years to build a superior schedule to the Stephensons,” said Trick. “To their credit, they were very, very good at what they did.”
Over those 10 years, Trick and his team developed underlying techniques to model the problem. Just as crucial, computing speed had vastly improved, which allowed for operations researchers to find faster, more complex solutions than was previously possible.
“Computers were faster, the underlying optimization code was faster, and something that used to take weeks would take minutes,” he said.
In 2005, Trick and his team took over the scheduling for Major League Baseball.
Bring in the data
Using optimization algorithms, Trick and his team initially used estimates and averages—called prescriptive analytics—to create the MLB schedules. But the era of faster computers had also ushered in a new tool: vast amounts of data. This data has been a treasure trove for operations researchers looking to build more accurate and predictive models.
For instance, by analyzing data from past games, Trick could determine different factors that affect attendance and, by extension, revenue. Is it better to schedule a baseball series from Monday through Wednesday, or Tuesday through Thursday? Analyzing the data could provide a concrete way to determine the schedule with the most potential for profit.
Trick’s team created the baseball schedule from 2005 and 2017. During this time, data was allowing operations researchers in every industry to ask new optimization questions. It was also raising questions about the quality and usefulness of data. Although an operations researcher may have access to enormous amounts of information, is it strong enough to make predictions?
A new era of scheduling
In 2018, MLB selected a new firm to build their schedule, one that could do it even faster, and in a way that was more responsive to the changing needs of the league.
For Trick, this development speaks to the future of the field of operations.
“I am an academic, so my approach has always been open,” said Trick. “The firm that won the contract in 2018 built upon what we were doing: they figured out a faster way to meet the MLB’s current needs.”
Trick warns about technological inertia, keeping the old objectives and constraints when circumstances are changing.
“In this field, technological capabilities are constantly improving and expectations and needs shift. Operations researchers must also adapt and shift to make sure they are addressing the most relevant, critical problems.”
As for Trick, he too is shifting his research focus. In 2018, he was part of a team that was awarded the 2018 Franz Edelman Award for Achievement in Advanced Analytics, Operations Research, and Management Science. The team created a revolutionary approach to meet the demand for the spectrum used for wireless communication in North America
“This was a great project to work on, and it was tremendously fun to be part of that team. There are huge opportunities for operations research as we integrate data and prediction to our work.”
Michael Trick is the Harry B. and James H. Higgins Professor of Operations Research at Carnegie Mellon University, and has served as dean of the Qatar campus since 2017. With undergraduate programs in biological sciences, business administration, computer science and information systems, more than 1000 students have graduated from CMU-Q since it opened in 2004.