AWS for M&E Blog
NFL and AWS Big Data Bowl will bridge the gap from college to the pros
Every April, National Football League (NFL) general managers face a question that has shaped franchises for decades: which college players will translate their game to the professional level? For all the scouting reports, game film, and gut instincts that go into the NFL Draft, the process remains one of the most complex in professional sports. Early-round picks and late-round picks alike can defy expectations in both directions. The data has always hinted that better answers might exist somewhere in the gap between a prospect’s college career and their first NFL snap. Now, the NFL and Amazon Web Services (AWS) are asking the data science community to go find them.
The ninth annual Big Data Bowl, the NFL’s flagship analytics competition, is taking a new direction. For the first time, participants will work with data that spans the full arc of a player’s journey into the NFL, combining college statistics, player tracking data from the NFL Scouting Combine, and Next Generation (Next Gen) Stats from their rookie seasons. The challenge is to find the connections that link what a player does before the league to what they do after they arrive. Join the 2027 Big Data Bowl and start exploring the data.
The 2027 challenge: Predicting professional performance
Previous editions of the Big Data Bowl have focused squarely on in-game NFL data, asking participants to evaluate rushing, passing, tackling, and defensive coverage using a single season of Next Gen Stats tracking data. This year’s competition builds on that foundation and takes it in a new direction.
“We’ve always focused on NFL stats, and this year’s Big Data Bowl will be no different,” says Mike Lopez, the NFL’s senior director of football data and analytics. “But the unique thing we’re going to try this year is to use what happens before players get to the league to understand what happens when they’re in the league.”
The dataset is built around players from the 2023, 2024, and 2025 NFL Draft classes at five position groups: wide receivers, tight ends, defensive backs, offensive linemen, and defensive linemen. For each player, participants will have access to tracking data (x-y coordinates of player movements) from both NFL Combine drills and NFL games, along with contextual information about the drills and plays in which they participate. The result is a dataset that follows a player from campus, through to the NFL Scouting Combine in Indianapolis, and onto an NFL field, giving participants the raw material to explore whether pre-draft data can predict professional performance in a way that is actionable and accessible to coaches, scouts, and front offices.
Drilling down, position by position
One of the defining features of this year’s Big Data Bowl is its emphasis on position-specific analysis. Rather than building broad, one-size-fits-all models, participants are being encouraged to go deep on a single position group.
“We don’t expect somebody to go in and say, ‘I’m going to analyze defensive backs and wide receivers and offensive linemen and defensive linemen, and I have an algorithm that’s going to work for all four of them,'” Lopez explains. “Football doesn’t work that way.”
At the Combine, each position group runs more than ten distinct drills designed by NFL position coaches. Wide receivers run route patterns. Defensive backs drop into coverage. Linemen work through change-of-direction and blocking drills. The movements are standardized, meaning every wide receiver in each draft class runs the same breaking route, told to do the exact same thing by the same coach. That consistency across years of data creates a controlled environment where meaningful comparisons between prospects become possible.
“We’ve always looked at that type of data with the 40-yard dash or the shuttle run time,” Lopez says. “But now the tracking data gives us something else. We can overlay all those wide receivers, figure out who’s good at it and who’s not, and then when players get to the NFL and run that same pass pattern, ask whether there’s something in the Combine that gives away how good they’ll be.”
The position-specific framing also opens the door to questions about scheme fit, a topic that has become central to modern NFL discourse. Different teams run different blocking schemes, different coverage packages, and different route trees. Coaches evaluating draft prospects are not only asking whether a player is talented. They’re asking whether that player fits what they do. The data this year could begin to quantify those scheme-level questions in ways that have previously existed only as coaching intuition.
Examples to consider
Your challenge is generating actionable, practical, and novel insights from Combine tracking data and exploring how drill movement translates to early NFL play. Examples include:
- Route running and coverage – Measuring change of direction, deceleration into breaks, and acceleration out of cuts for receivers and defensive backs compared to in-game separation
- Trench play – Evaluating first-step quickness, lateral agility, and burst for offensive and defensive linemen during agility drills compared to pass-rush or run-blocking performance
- Drill translation – Comparing sensor-based movement metrics with traditional drill times to identify where tracking data adds context to prospect evaluation
- Position-specific mechanics – Analyzing body control, acceleration profiles, or pursuit angles during position drills and their relationship to NFL rookie performance
This list isn’t exhaustive, and we encourage participants to be creative with their submissions.
Finding the signal in the scouting process
The motivation behind this year’s challenge is rooted in a well-documented reality: predicting NFL success from the draft is uniquely difficult compared to other sports. In the NBA, NHL, and Major League Baseball, the relationship between draft position and professional performance is steeper. Players selected earlier tend to outperform those selected later by a more reliable margin. In the NFL, that relationship is less pronounced, reflecting just how hard the projection is. Draft position and professional performance align less tightly, and the margin between them is harder to predict.
“We’ve had, for years, really decades, players slip through the cracks,” Lopez says. “That leads to some level of inefficiency where players drafted earlier aren’t always the better players or who have the better fit in the league. If there is something in this tracking data that can help teams make smarter decisions in their draft process, that would certainly be something they’d be interested in.”
A good example that illustrates this point is NFL player Puka Nacua. The Los Angeles Rams wide receiver, selected in the fifth round of the 2023 NFL Draft out of Brigham Young University (BYU), went on to break the NFL’s rookie receiving records. His Combine data and college production were available to every team. The question is whether the right analytical framework, applied to the right data, could help surface that kind of signal earlier in the evaluation process.
From competition code to game day
One of the distinguishing features of the Big Data Bowl is the speed at which winning solutions become real tools. Thanks to the AWS data science team’s work with Next Gen Stats, the pipeline from competition submission to on-air deployment is remarkably short. Past competitions have produced more than 75 machine learning (ML) based metrics, including Rush Yards Over Expectation, Tackle Probability, and Pressure Probability, all of which appear on NFL broadcasts.
“The winning Big Data Bowl algorithms don’t stop on Kaggle,” Lopez says. “They are used by Next Gen Stats. They’re used within teams. They’re used in versions of public NFL research over and over again.”
The AWS data science team plays a critical role in that pipeline, taking competition code and scaling it for real-time production. “We read through it, we port over the code, and the AWS data science team has scaled a lot of that,” Lopez explains. “It’s made it faster. It’s made it seamless to get that information back in near real time, which helps our broadcasters get smarter, helps our fans get smarter, and after the game, when it’s shared with teams, helps our teams get smarter.”
This year, the most natural home for winning solutions is Draft IQ, the NFL’s suite of Combine and draft analytics tools. A standout model from the competition could be integrated into Draft IQ and put into production in time for the 2027 NFL Draft, giving participants the opportunity to say their work directly shaped how the league evaluates its next generation of players.
Exploring uncharted territory
Lopez is candid about one aspect of this year’s competition that makes it different from previous editions. Because the dataset is fundamentally new, and because the question of whether Combine tracking data predicts NFL performance has never been systematically explored at this scale, there is a real possibility that some analyses will return null results. And that is fine.
“I have a feeling a lot of people are going to look at things and think, ‘Nope, that doesn’t predict anything,'” Lopez admits. “We have to be prepared for some null solutions. We don’t know what’s going to work, and I think that’s one of the exciting things.”
He’s emphatic that participants shouldn’t feel pressure to force a finding. “What I don’t want people to do is try something, not find that it’s predictive, and then feel like they have to tweak the answers to find the right group or the right player that it fits with. Just be okay with it and focus on the process and the idea more than the strength of the relationship.”
In some ways, a well-executed null finding is as valuable as a breakthrough. If the data shows that a particular drill doesn’t meaningfully predict NFL performance, that saves teams time and resources they might otherwise spend chasing a false signal. The competition is as much about establishing what matters as it is about confirming what doesn’t.
A community that changed the game
The Big Data Bowl has grown from an experiment into one of the most influential events in sports analytics. More than 100 participants have gone on to jobs with NFL teams, the league office, and other professional sports organizations. The competition’s collaborative community, where participants openly share code, exchange ideas, and sometimes combine teams mid-competition, has created a talent pipeline unlike anything else in the industry.
“It’s the most positive, engaged sports analytics community I’ve ever seen,” Lopez says. “People are here because they love football and they want to push the game forward.”
As the competition enters its ninth year, that community will get its most distinctive challenge yet. The data is new, the questions are open-ended, and the potential applications extend from broadcast graphics to front-office draft boards. For data scientists, developers, and football enthusiasts, it’s a rare chance to work with authentic NFL data and potentially reshape how the league discovers its next generation of talent.
Registration for the 2027 Big Data Bowl opens this fall. Have questions or ideas? Share your thoughts in the comments or connect with the Big Data Bowl community.
Further reading
https://aws.amazon.com/sports/nfl/
https://operations.nfl.com/programs-initiatives/innovation/big-data-bowl