NFL Big Data Bowl Experience
Intro
These last few months my teammate Andrew Lins and I decided to compete in the 2021 NFL Big Data Bowl. Neither of us had any experience doing data science competitions, but we both had some data science experience and a lot football experience. We knew that we would be facing tough competition, so we took a different approach: we wanted to emphasize a solution and use case not simply a report.
Thesis
The NFL is a passing league now, but from a defensive standpoint big plays do not tell the full story. Maybe you blow a touchdown, but what if that was the only reception you gave up all game? Maybe you don’t have that many picks this season but you’re holding receivers to under 50 yards 95% of the time like Jalen Ramsey. Banking on the big play factor becomes less valuable: According to Pro Football Reference the percentage of passes being completed has increased on average each year while the percentage of interceptions has decreased. And as the game gets faster and the defensive line positions have adapted while the offensive line has stagnated, we have QBs compensating by neutralizing explosive defensive linemen with quick releases. We are at a time where secondary units can be very dominant and criminally underrated like the Buffalo Bills. Players like Kendall Fuller and Adrian Amos are not getting the credit they deserve until recently. Teams need to focus less on breakups and interceptions and more on yards given up, first downs given up, and if they take opposing receivers out of the equation.
The App
We wanted to make a tool based on this 2018 data that could display the insights from our notebook in the most clear and useful way. This is why we converted our data frames to json and also cross referenced the players in our lists with their 2018 contract info from spotrac. With this data we created the web app #MoneyTeam to view the top players by our seven analyses. We ran into a few issues uploading the notebook to Kaggle, and so we are very glad to have made the app before as well.
The most exciting part about these lists was seeing our rookie contract list and how well these players have panned out 2 seasons later. Our entire rookie contract top 10 list made the playoffs this year. How well these players have panned out makes sense because our lists reward the fundamentals and those are consistent. Our system will be improved on but we see this app as a very useful tool to acquire or make a play at sleeper players. For example, Adrian Amos was on the Bears in this 2018 season, two years later he’s on the Packers and they have one of the best secondaries in the league.
It does not make sense for teams to weigh interceptions heavily. Micah Hyde is a great player on a great secondary but to traditional stat sheets he does not stand out, Jordan Poyer and John Johnson are underrated, and then you have Jalen Ramsey who only had 1 interception this year but wouldn’t give up more than 50 yards to anyone less than the duo of Aaron Rodgers and Devante Adams. In fact he was such a dynamic player he essentially played OLB 40% of the time this year. #MoneyTeam weighs that more and teams should build their teams around that. Fundamentals carry you and are consistent. It’s no coincidence that our rookie contract list is holding up.
I am very confident that this app is onto something and with continued focus and refinement it could be useful in any front office.
Conclusion
One of the best decisions I made in quarantine was deciding to do the Data Bowl because it forced me to get creative, learn new skills, and improve on old ones. It seemed so intimidating on the outside seeing submissions from last year but then we talked about it and said: “They used what they know to submit what they had, we have our own skills and experiences”. From there we built something unique. It was exciting seeing that our system holds up well and think it will continue to. The only question for us now is where we take #MoneyTeam from here after the competition is over.
