Using data science tools to automate processes and advance knowledge of the sport of american football. I employ Python-based libraries to collect and manipulate data, build predictive models and uncover novel insights.
In this study we identified and classified pass stunts from Weeks 1-8 of the 2021 NFL season using player tracking data. We use these observations to test commonly held beliefs about successful protection against stunts and uncover other relevant aspects using feature importance measures. Model probability outputs are then used to distinguish likely pass rush wins from likely losses and compare feature expression between these groups. We then model pass rush win rate allowed using predictions from an LSTM classifier and discussed applications to player evaluation in opponent, self and pro scout settings.
Scrapes and parses college football play-by-play data using BeautifulSoup4. Contains modules for storing data with SQLite and producing csv spreadsheets that can be imported directly into Hudl.
Uses selenium to scrape interactive web data from three major football recruiting services. Employs fuzzy matching and other distance metrics to generate comparison scores, which are used to facilitate manual selection of matches.
Python program using csv exports of charted Hudl game film to generate custom opponent tendency reports.
Charts, graphs and animations created with matplotlib and seabornfor the Stunt Busters report.
