Football

This project was assembled as part of Nathan Wright working as an analyst for Colonial High School Football. In this project, the results of numerous works of research in playcalling, decision-making, outcome modeling, and defensive expectancies were merged together to design the statistically optimal offense. If you're going to read any project on this website, this project is the one. 

This project introduced xWEPA, a framework for quantifying quarterback decision-making in the passing game. By combining expected play value (xEPA) with quarterback-specific completion probabilities (qbxcp), the metric isolates the quality of the decision itself rather than its realized outcome. Results showed that xWEPA retains predictive value for wins and scoring, significantly reduces variance compared to EPA and WEPA, and strongly aligns with performance in high-leverage plays. These properties position xWEPA as a process-based complement to traditional efficiency metrics, offering new insights for coaching, scouting, and roster evaluation.

This project engineered a situationally aware framework to visualize expected defensive responses—rush count, coverage type, and adjustments—based on down, distance, motion, and play intent. From there, machine learning models analyzed pre- and post-snap alignment data to surface trends in defensive behavior, enabling predictive insights for offensive strategy and game planning

This project analyzed NFL player tracking and contextual data using Python to investigate whether wide receivers reveal play intent through pre-snap movement speed. Statistical models and visualizations were then built to identify significant tempo differences in targeted vs non-targeted WRs, offering actionable insights for scouting and defensive game planning.

A bespoke project, this report conducted full-scope opponent scouting report on the Arizona Cardinals using publicly available NFL data (nfl_data_py), analyzing offensive/defensive tendencies, situational behavior, and individual player profiles. I then created statistical visuals and actionable insights for game planning, including coverage-based QB performance, blitz patterns by formation, route-specific defensive efficiency, and fourth-down decision trends.

This is a pairing of two projects: a study on 4th down conversion attempt success rates & a 4th down decision-making dashboard. In the initial analysis, success probabilities of conversion attempts based on a variety of situational, contextual, and player variables were examined to predict whether a conversion attempt will be successful at an 86% success rate. Subsequently, a decision-making dashboard and framework was made to visualize expected gains based on situational inputs, and output a recommendation based on the highest Weighted Expected Points Added and expected Win Probability Added.

This project used play-by-play data to model epa and completion probabilities of routes when targeted in the passing game and display "Weighted Points Added", an output of an expected gain calculation for the expected points added (epa) upon a completion x completion probability (cp). 

This study analyzed 2021-2023 NFL play-by-play data using Python (pandas, matplotlib, seaborn) to evaluate early-down playcalling strategies and identify trends in 3rd-down distance-to-go & conversion success rates across all 32 teams. It conducted data preprocessing, statistical analysis, and data visualization to deliver actionable insights on playcalling tendencies, optimizing offensive efficiency for professional-level decision-making.

This study is a combination of a study on ypc when attacking certain gaps in blocking schemes along with an expansion into how YPC changes with formation and personnel deployment. The second of which was developed in a scouting report style, catered to the reading of a football coach.

This study built a model evaluating red zone playcall efficiency, focusing on down, distance, field position, and defensive alignment. From there regression analysis was conducted to quantify expected points added (EPA) for various offensive concepts in red zone scenarios.  

This study analyzed NFL salary cap allocations and draft capital with Python, identifying positional spending trends and performance correlations using OverTheCap and Sports Reference data. It then developed actionable insights on cap management strategies to optimize team-building, aligning financial decisions with player impact and organizational success. 

This study analyzed data from first and second-round NFL draft picks (2013-2020) to identify correlations between college performance and NFL success. Python used for web scraping, data collection, cleaning, and feature selection. Developed machine learning models (Logistic Regression, Random Forest) to predict NFL career success based on college statistics, awards, and other factors. Applied cross-validation and hyperparameter tuning to optimize model accuracy. It then provided recommendations for NFL teams to enhance draft strategies using composite metrics, with broader implications for data-driven decision-making in sports analytics.