How the NFL Leans On AWS Cloud Computing to Create the Regular Season Schedule
Each year, after the Super Bowl, the NFL has roughly four months to create the schedule for the upcoming season. The NFL has to account for 272 games, where each team (32 total) plays 17 games (new for this year) during an 18-week period, while optimizing for fairness, travel, and availability of the venue.
Prior to utilizing AWS cloud computing solutions, the NFL used to ship servers to their data centers and create “rack and stack” rooms of servers that would require manual configuration and tuning for optimal schedule outputs. Today, the NFL leverages AWS cloud computing to use predictive analytics for optimizing ratings (primetime Thursday, Monday and Sunday afternoon), relaxation of single-header protection rule, and the continuation of flex and fold policy as well as the adjustment of divisional weekend Sunday kickoff times.
Looking ahead, the NFL intends to leverage AWS cloud computing to use predictive analytics for optimizing ratings (primetime Thursday, Monday and Sunday afternoon), relaxation of single header protection rule, and the continuation of flex and fold policy as well as the adjustment of playoff weekend kickoff times
The NFL has built a new schedule scenario every day since January 4, 2021, adding new rules and constraints each day in an effort to better inform the search and arrive at their best possible result. They’ve analyzed roughly 80,000+ finished schedules, and they’re not done yet.
There are nearly 1 billion options for each team and hundreds of trillions of potential completed NFL schedules overall. AWS allows the NFL to find and analyze thousands of schedules each year that they wouldn’t have seen otherwise and answer questions like “what would it cost the NFL to change this?” or “what is the impact of this change?” so they can make smarter decisions and presumably arrive at a better final product.
Some factors that are considered for developing the complex schedules for each team include:
- There are more than 100 potential stadium conflicts (MLB, MLS, NCAA, high school sports, concerts)
- Minimize the number of teams that play a 3-game road trip
- Minimize the number of teams that play a road game following a road Monday Night Football game
- Minimize the number of teams that play 2 road games to start season
- Minimize the number of teams that play 2 road games to finish a season
- No team plays consecutive road games involving cross-country trips unless requested
- Teams can travel no more than 2 time zones for the Thursday game
- All teams playing home Thursday games have limited travel the previous week
- Two teams must play Thursday of Week 13 after playing Thanksgiving Day
- Minimize teams that play multiple road games against teams coming off their BYE
- Accommodate club requests/willingness to host holiday games on Thanksgiving and Christmas
- Division series should be well-spaced and spread throughout the season.
- Maximize separation between teams’ BYE week and Thursday game (mini BYE)
- Equal number of games “cross-flex’d” between CBS and FOX
- Strong Sunday, Monday, and Thursday Night Football Schedule
- Maximize late-season division games
- Minimize early season 1 p.m. games for teams with weather concerns
- Covid-related rescheduling of games and locations
- Addition of a 17th regular season game for the 2021 season
The addition of the extra game has both a positive and not-so-positive impact. Rhe positive is that those additional games are “free agents” and don’t “belong” to CBS or FOX based on the road team like the other 272 games. That flexibility helps the NFL as they try to construct quality Sunday afternoon windows while also trying to put good matchups in all the national/primetime windows. The downside to the extra games/week is that the solution space that was already essentially infinite didn’t just double or triple in size, but increased exponentially, so it takes the NFL a little longer to search through the void and find contenders.
AWS allows the NFL to access thousands of machines on a daily, nightly, or hourly basis letting them to think bigger, analyze more, and react to things.
The NFL not only relies on AWS’ elastic compute capacity for 1000s of instances to get their scheduling complete but they are able to scale their capacity using 80-90% Spot Instances. Amazon EC2 Spot instances allow the NFL to use EC2 spare capacity when scaling up their scheduling jobs optimizing the cost to complete the daily scheduling runs.