NFL Schedule, Rivalries and Potential TV Ratings Optimized by Packaged Software
The NFL's 2008 season has kicked off, and specialized optimization software helped league execs wade through thousands of scheduling variables to come up with the best lineup for teams, TV networks and fans.
By Thomas Wailgum
The National Football League is a professional sports organization known for its meticulous, hands-on approach to everything—how the league contracts with TV networks, how teams draft their players and how those players should act on and off the field, how licensing deals are signed, and how rules are enforced on the playing field.
The same was true with how NFL executives created the schedule for its teams every year: It was all done by hand, starting the day after the Super Bowl, with a peg board and little tags. “The process was kind of secretive,” says Michael North, the NFL’s director of broadcast planning and scheduling, who’s been with the league for 15 years. “We would go into the room, lock the door and emerge 10 weeks later: ‘Here’s your schedule. Play it.'”
That all changed several years ago, when the NFL realized it could use technology to automate the process—making it more efficient and its schedules better—and a Canadian manufacturing engineer named Rick Stone came knocking on the NFL’s door.
Stone, who had done sports scheduling for several minor hockey leagues in Canada, got his shot at the big time when the NFL was building its 2004 schedule. Stone and his optimization software scored. “It’s pretty hard to compare a couple of guys in a room trying to put this jigsaw puzzle together to what a computer could do,” Stone says. “It’s not really a fair fight.”
North and the NFL brass have been duly impressed. “If you’re building the schedule by hand, you box yourself in real quickly and you start making compromises,” says North. “More than anything, Rick has developed a tool to prioritize, in any given season, certain factors and constraints that lets us figure out which of these schedules are delivering as close to optimal as we can get.”
On Any Given Sunday, or Monday, or Thursday…
To the average football fan, the task of creating each season’s schedule might not seem that complex. (The 2008 season started on Sept. 4.) After all, there is a fixed system and planned rotation in place that, come the day after the Super Bowl each year, lets each of the 32 teams know who their opponents are going to be next season. Therefore, it’s just a matter of determining when those 256 NFL games will occur. (To read about the security of quarterback-to-coach communications at NFL games, see “How Secure Is All That Wireless Equipment at the Super Bowl?”)
But just as the league’s prominence has risen to the top of the American sports scene, so too has the number complicating factors that can affect the schedule and the teams playing the games. In addition to television networks (including the NFL’s own cable network) that are spending billions on broadcasting rights and want the very best matchups, all of the teams want to avoid schedules with horrific travel arrangements. And then there’s the demanding, ticket-paying fan base who want to see only the best games and teams come to town.
“So we’ve got to figure out how to put that 256-game jigsaw puzzle together in a way that’s competitively fair to the 32 clubs,” says North.
Here are just a handful of the factors that have to be considered: whether teams should play on Thursday, Sunday or Monday nights, or as part of Sunday doubleheaders or on certain Saturdays; how best to avoid three-game road trips for teams or road games after road Monday night games; watching out for consecutive seasons where teams get the earliest possible bye week; determining who certain teams play first and last; which games warrant prime-time TV exposure during sweeps weeks; which teams should play on the national TV schedule and which teams shouldn’t; which teams play in “specialty games,” like on Thanksgiving day or on other holidays; and climate issues, such as which teams go to Miami in the early part of the season and Green Bay, Wis., late in the season.
“Any factor you can think of has to be factored into this decision,” North says.
There are even things outside the NFL’s control that have to be considered when making the schedules, like the likelihood of hurricanes and other sports teams’ schedules. For example, North says 11 NFL teams share stadiums or downtown geographic areas with major league baseball teams (the football season overlaps with the baseball season, and then there’s the playoffs and World Series to contend with). “So you have to work around all those teams in September,” North says, “and you need to be able to adapt to post-season conflicts that arise.”
There are, in fact, some 6,000 of these factors that the NFL prefers to have included in the creation of the 256-game schedule, Stone says. So, just how complex is process? “Say you’re going to buy a lottery ticket, and you’re trying to pick six numbers out of 50,” Stone says. “That’s like 15 million combinations, or something like that. So if picking six out of 50 is so difficult, imagine the number of combinations of picking 256 out of 6,000. It’s astronomical.”
What’s most amazing, in retrospect, is that the NFL, with around $7 billion in annual revenues, had been able to do this by hand for so long and still come up with a pretty fair schedule.
“Frankly, it’s staggering that the National Football League, a multibillion-dollar company, was taking arguably its most important asset—the schedule—and building it by hand only about 10 years ago,” North says. “It’s remarkable to think where we were not that long ago.”
Cold Calling the NFL
Rick Stone, a manufacturing and industrial engineer based in British Columbia, was still working his scheduling job when he launched a fledgling software scheduling business for professional sports, called Optimal Planning Solutions. “It’s the exact same technology just applied to a different field,” he says. At a high level, Stone says that his software creates a mathematical model that incorporates all the various scheduling constraints, rules or requests—as in “like to have” or “must have”—and then it uses a commercial scheduling application to solve the mathematical model. Thus, an optimal schedule is created.
So, as Stone tells it, he cold called the NFL at its stately Park Avenue offices in New York City. “I was already doing sports scheduling for probably about 10 leagues at the time around the world, mostly minor league hockey and lower-tier leagues,” Stone says. “And I basically called them out of the blue and said ‘Hey, this is who we are and would you be interested in our software?'” (The “we” was actually “he,” as Optimal Solutions was a one-man operation at the time; it’s up to two people now.)
North recalls Stone’s phone call and that Stone had done work for the National Rugby League of Australia, “which happens to be structured a lot like the NFL,” North says. “Everybody plays one game a week, you play division opponents home and away, and there’s one key game a week.” The NFL had initially put out an RFP looking for a technology partner in 2001. “We talked to everybody we could think of—public and private, to Carnegie Mellon and MIT to NASA,” North says. “We talked to other professional sports schedulers that were making schedules for the NBA, Major League Baseball and the NHL.”
The NFL had selected another company’s scheduling software, which it used for the 2002 and 2003 schedules. But there was something about Stone’s software. North says Stone worked in a “on spec” role in 2003. “We started slowly with Rick, and he worked in parallel with us in a pilot,” North says. “OK, guy in your basement in Vancouver, you can take a shot over there on your own time. In the meantime, we’re going to keep doing it the way we were doing it until you prove to us that you’ve got a tool that can do it better.”
The result? “Very, very quickly he was able to prove to us that his software was very capable,” North says.
Stone says the schedule his software came up with was “substantially different” than what the NFL came up with that year. “They ended up using our schedule,” Stone says. North says Stone was hired as the NFL’s primary solution provider in 2004.
Safety in Numbers
The NFL now has the Optimal Planning Solutions software onsite. North says that while they still take roughly the same time to make the schedule—10 weeks, from the day after the Super Bowl until the first week in April—”we work so much smarter, so much more efficiently,” he says. “We can see the opportunity costs of some of our decisions. Whereas in the old days, once you put Cincinnati and Cleveland in week five, that decision was made. You have to just go down that path.”
Now, North says, they can finish a thousand schedules with a Cincinnati and Cleveland game in week five, and then, if need be, put in a rule that says: Cincinnati and Cleveland can’t be in week five. “Are any of those better, and at what cost to the rest of the schedule?” he asks. “You can have this level of analysis and comparison that we were never able to generate before.”
Each schedule generated, taking into account the parameters and constraints set up, receives a score. For example, if one network gets the game that it wants, that schedule gets positive scores. If there’s a situation where a team has to play three games in a row on the road or a short week: playing on a Monday, followed by a Sunday game and both are road games—those are negative things that count against the schedule.
“We can enumerate everything that goes into a schedule, and then we try to get the best possible score,” Stone says. “And that scoring system continuously changes as you’re trying to marry the request of the fans versus the networks versus teams versus everything else. There’s always changing priorities as you’re going through this.”
What’s important to the NFL is that the software hasn’t become the final decision-maker; the software facilitates and enhances the decision making as the NFL strives to better serve its 32 teams, TV network partners and fans.
“The beauty of the software is that we can go into every single game, every single week, every single team, every single place where we could be playing a game and we put some parameters around it,” North says. “You can say to the software: Mix and match these any way you want to, just know that each game comes with a certain set of parameters. I don’t want to play this game any earlier than week three or this game any later than week 14. That game should fall in sweeps week, somewhere between 9 and 12. That game should be in Pittsburgh instead of in Cleveland. That game needs to be on a Thursday night, which means that both teams need to be playing a home game the week prior.”
“The software,” he adds, “can do what we were only dreaming of doing 10 years ago.”
Trillions of Possibilities and Big Benefits
The first and perhaps most important benefit of the software is that North says the NFL is able to achieve “competitive fairness” in its schedules every year. “There are literally trillions of possible NFL schedules every year,” North says. (Stone thinks North’s estimate is low. The possibilities are “infinite,” he says.) “Now we get to look through thousands and thousands and thousands of them and pick the ones that are fairest overall to everybody as opposed be being able to get one done.”
Since they’ve been using the software, NFL execs have been able to uncover scheduling trends that potentially could have placed one team at a disadvantage, North says. For example, data showed that it was unfair for teams that had to travel to play another team that had had a bye week the week right before. “That was something we were never able to account for or even think about up until about a couple of years ago,” North says. “We were doing it without even knowing.”
So, with the Optimal Planning software, the NFL has built a constraint into the application which ensures that that won’t happen to any team more than twice in a season, and only once against a division rival (each division has four teams). “Little things like that improve the overall quality of schedule every year,” North says.
North says that the quality of the schedules and the ability to score each schedule—”instead of taking a subjective view, having an objective view,” he says—have contributed to the increase in TV rights payments by the networks. “These networks are playing a lot of money to the NFL. If they weren’t being given quality schedules, if they weren’t able to reach the most amount of fans and satisfy sponsors and advertisers, then they wouldn’t pay those rights fees contract after contract, season after season,” North says. “I can’t help but think the scheduling process, becoming more automated and more efficient, has enabled the league to deliver a higher quality product to our business partners.”
But this is only the very beginning, North points out. “Initially it was: Holy cow, look we can get a schedule done with a computer. That’s pretty neat,” he says. “Now it’s, OK, here’s a million more things that we want to be able to do.”