The NFL's schedule makers face a lot of uncertainty when they sit down every spring to put together the next season's Monday Night Football schedule. They want viewers and they want to give teams national exposure.
Yet the games won't be played for several months, and all sorts of things can happen that make a game which seems compelling in April a viewership bust in October: Players might be traded, injured or leave in free agency, and there is often a great deal of parity so a playoff team from one season may take a nosedive the next. A Monday Night Football schedule that once appeared to be filled with ratings winners is riddled with games few fans are interested in watching.
Scholars from the University of Iowa's Tippie College of Business say they may have found a new game plan to improve the probability of exciting Monday night matchups. After analyzing years of Monday Night Football viewership, the scholars developed an optimization model they believe shows its schedules could be improved. Bhupesh Shetty, a management sciences doctoral student; Jeffrey Ohlmann, associate professor of management sciences; and Gary Gaeth, professor of marketing, analyzed every Monday Night Football game played from 1993 to 2008 and broke each game down into its various factors to see what has the biggest impact on a game's Nielsen ratings.
Among the factors considered were the teams' records, the date of the game, competing sports events on other networks, and the quarterbacks. But many of those variables that are strongest predictors of high ratings (the records of the teams playing, for instance) are unknown in April. So Shetty, Ohlmann, and Gaeth teased out the factors that are known before the season begins and could be helpful when putting the schedule together in the spring.
Those factors the study identified as having the highest impact:
- Games played by the defending Super Bowl champions
- A team with a high-profile personnel change, such as a star player or well-known head coach joining the team
- Teams with high-powered offenses, or, as Ohlmann says, "defense might win championships, but offense draws the ratings."
"And for a chance at a late-season ratings king, you can't go wrong scheduling two top quarterbacks against each other," Ohlmann says.
The analysis also discovered that certain teams generate better ratings even after controlling for other factors such as team record, superstar quarterbacks, etc. The Dallas Cowboys are the top draw, followed by the Minnesota Vikings, Pittsburgh Steelers, and Green Bay Packers.
One of the strongest variables that determines a game's ratings is the presence of the Jacksonville Jaguars. The study found that whenever the Jaguars play, the ratings consistently tank. This was true during the entire sample size dating back to 1993, the year the Jaguars joined the NFL as an expansion franchise, so it includes the period when the team won division championships and regularly made the playoffs.
So fans consistently avoid watching the Jaguars. Understandable for a small market team that got an expansion basically to stop other teams from threatening to move there and hold their existing cities hostage.
Like everyone in academia who makes models, they created it to predict the past and hope that means it might predict the future. They compared their optimized Monday Night Football schedules for 2011, 2012, and 2013 to the actual Monday Night Football games played in those seasons. It turns out that their optimized schedules were expected to have produced about 10 million additional viewers per season than the actual games scheduled. Ohlmann concedes that these gains are not provable, however, given the level of viewership variability and the fact that their optimized schedules are hypothetical.
Still, the optimized schedules avoid games that actually resulted in ratings busts and provide a tool for broadcasters like ESPN as they negotiate which games they can televise.
Gaeth's and Ohlmann's paper, "Analysis of Monday Night Football viewership," is published in a recent issue of the journal Statistical Analysis and Data Mining.
Comments