New York Giants cornerback Ross Cockrell (37) defends against Detroit Lions wide receiver Kenny Golladay (19) on an incomplete pass in the first half during a NFL football game at MetLife Stadium.

Robert Deutsch - USA TODAY Sports

UB Professor Working with NFL to Create Better Schedules

"There are more possible arrangements...than there are atoms in the universe."

April 18, 2019 - 3:39 am

BUFFALO (WBEN - Brendan Keany) - Operations research is defined as "the application of scientific principles to business management, providing a quantitative basis for complex decisions."

"Any decision problem, you can use these mathematical models," said Mark Karwan, a professor of operations research at UB. "You have to come up with the models, and you have to leverage ways to solve them by searching lots of possibilities through trees and something called branch and bound."

Karwan has solved problems before that include real time energy pricing and new facility locations - problems with tens of thousands of constraints and variables.

"Those could be solved in four to five hours on a high speed computer," he said.

However, Karwan is now working on a new problem that requires an unfathomable amount of data - the NFL schedule.

RELATEDBills Schedule Released

"Take the Bills," he began. "They could play any of their 16 games in the first week, any of their 15 games in the second week...that's 16 times 15 possibilities...14, 13, 12. Do that on your calculator and it's over 20 trillion just for one team. That's not counting if it's Sunday, Monday or Thursday that week, so that's over 60 trillion. When you do all teams, there are more possible arrangements of the 17 weeks than there are atoms in the universe."

Karwan became involved in processing NFL scheduling outcomes several years ago when a student approached him about doing an independent study in the field, but the student wanted the project to be about sports in some capacity. A few days later, then general manager Russ Brandon complained that the Bills schedule lacked fairness because the team had to face numerous teams who were coming off their bye week. That got the pair thinking, and they began brainstorming models to create a more fair NFL scheduling system.

Listen to the full story and interview with Karwan below:

Essentially, they looked at variables such as consecutive road games, start times, travel, bye weeks, etc., and they compiled constraints and manipulated certain variables to make a season as fair and balanced to every team as possible. Their models were a success, and they submitted a paper which they got to present at a sports analytics conference at MIT.

"We don't just want a feasible schedule; we want want one that minimizes the number of teams that have three-game road trips in a season," said Karwan. "But if you had one last year, then we don't want you to have one this year, so we'll put a 100-point penalty and some objective. Then, we try to minimize all of these penalty points, subject to all of the conditions that make a feasible schedule."

And their presentation apparently made some noise.

"One of the NFL media people called UB's media person and asked if I would like to speak with their head of scheduling, Mike North," continued Karwan. 

North was intrigued with the work that Karwan's team was doing, so he shared information with the team that represents the main cause for headaches in the NFL scheduling process, and one that Karwan's models didn't initially consider - venue conflicts.

Karwan says that 40-percent of all Thursday, Sunday and Monday location days are blocked due to other engagements.

"They have Taylor Swift concerts, NASCAR in North Carolina, college bowl games at the end of the year," said Karwan. "[North> saw what he were trying; he saw our backgrounds, and he said, 'How about I send you the stadium availability data?'"

Karwan's team attempted their model sets with the vast new constraints of the stadium availability data, and it really presented a major problem at first. However, North stuck with them and continued to reach out and engage the team on the process, and he even sent the team much of the software the league uses to develop and generate the schedules.

Now, Karwan's team is partnering with the league in an experimental capacity, essentially plugging in models and constraints to try new ideas and test rule changes.

"Do we really need to have a divisional game the last game of the season? Do we really need a bye week after an international game when London to New York is no worse than London to New York to San Francisco and you have a time advantage coming back from London?" he asked. "They're starting to manage those with the owners, but there are still things that are required or requested that may be hurting the ability to have a much lower penalty. It's hard to say what things affects it the most. As stadiums get more and more blocks, we know that's going to make it harder and harder."

While Karwan's team will have to wait until the schedule planning period for the 2020 season to be directly involved in the process, he says the team can take this season to further develop their models and make tweaks based on the 2019 schedule.

"We have until next planning season to further develop our method and see if it would have helped 2019," Karwan added. "So going into the 2020 season, we would have a horse race where we try to do this with them and see if, at the end, who gets the best schedule, or at least very comparable schedules, and they will choose."

Comments ()