Tackling NFL Defensive Data with R Shiny
Introduction
The purpose of this application is to showcase the important information on defensive players to the audience that needs it. This audience can include defensive player scouts, NFL team owners, fantasy team owners, NFL fans, or even those interested in learning more about the defensive player base. I used R to develop the visualizations, data cleaning and R Shiny to deploy the application.
Click here to jump into the application.
Data
The datasets are from the 2024 NFL Big Data Bowl from Kaggle. They add up to 13 CSV files that take up 1.61 GB. This was an issue for developing the R Shiny app due to its limit of 1 GB for performance. That is the reason why most of the project was spent optimizing and reducing this data down to the essentials needed to deploy successfully. (To see the original cleaning file, you can contact me) Below is a sample from one of the CSV files.
This project was a great source for anyone looking to optimize a large application into smaller requirements as well as utilizing joins in R. This can translate directly into SQL experience which is sought after in the data analysis space.
Player Tracker
The player tracker is made for the user to evaluate individual players. There are two charts available, the first is a bar chart with a trend line and the second is a radar plot. The first chart matches the selected player’s total tackles for each week against the top 10 tacklers of that given week in their position. The radar plot displays six attributes of the selected player alongside the ideal metrics for their position.
Both are shown below.
A good example of a use case in this situation is selecting Quinnen Williams, who has proven to be an almost perfect match to the benchmark and is rated as one of the best in his position. He received a four-year, $96 million contract extension by the New York Jets, and he became the second highest paid defensive tackle in the league at the time.
Team Tracker
The team tracker is made to assess the current strengths and weaknesses in a team’s defense. A user can look at teams that have a plethora of highly rated players in a position to shop for trade deals.
A good example of a use case for this application would be a player injury of someone who is critical to defense. If such a player needed to be replaced for the team to stay competitive, they could look at teams that have X amount of players in the same position. Then if their overall rating is favorable, they would likely take a deal to capitalize on value.
Leaderboards
The leaderboards are shown to give users a perspective on who is leading in selected weeks or week spans to show a benchmark of the top 10 in a given stat. Users can select from various stats, such as Assisted Tackles, Impact Plays, Tackles, etc. to see who is excelling in those areas.
This can give a more in depth view on a specific area that the top players in a stat-group excel at. I also included some negative player stats, such as missed tackles to include a performance check on both sides of the coin.
Live Play Animations
The live play feature of the application allows users to select a play and watch it rendered as an animation with different colors for the teams. The football is also a unique color to show the focus of the play and how tacklers are condensing on the ball carrier. You can hover over the points on the field to view the name of an individual player.
Limitations / Future Work
The limitations of the data provided could have hampered the ability to accurately capture the rating and importance of defensive players. Certain events or occurrences, such as interceptions, passes deflected/tipped, blocks overcome, etc., weren’t captured in the data set. The data was also only for the 2022-2023 season in weeks 1-9. Since it was centered around a project for tackling metrics, the data didn’t provide important offensive metrics.
For future work, I believe it would be immensely powerful to include a free agency, contract value, and market value analysis of players. The ability to incorporate a ‘Moneyball’ approach into football in evaluating players based on cap space would be extremely influential to owners and agents.
I would also like to do an analysis of their college statistics and competition to provide projections to how they would do in the NFL.
With the increasing data and advancements in the AR/VR industry, I believe the animation in live plays could eventually be transformed into a virtual reality where you get to watch the best plays over time in live action as if you were on the sidelines.
Links
Github: github.com/briangdrewes/nflapp
Google Slides: NFL Big Data Bowl 2024 Defensive Scout App Presentation
LinkedIn: Brian Drewes
Application: Defensive Stats App
Attribution:
Featured Image by master1305 on Freepik
About Author
