Model down, distance, score, field position, and overtime rules for decision support.
Football analytics
Football
Drive outcomes, coverage tendencies, win probability, and roster evaluation research.
Workstreams
What this team studies
Study coverage, spacing, and drive context behind high-leverage defensive outcomes.
Build repeatable frameworks for roster, prospect, and skill-position evaluation.
Featured work
NFL OT 4th Down Decision Engine
A decision-support tool that recommends whether NFL coaches should go for it, punt, or kick a field goal on 4th down in overtime — trained on 350,000+ plays from 2016–2024.
Problem
NFL coaches lack a real-time, data-driven framework for 4th-down decisions in overtime, where post-2022 rules guarantee both teams a possession before sudden death.
Approach
Four XGBoost submodels — punt outcome, field goal probability, conversion probability, and win probability — are chained together with isotonic calibration and expected-value maximization to recommend the optimal decision given field position, score, team quality, weather, and venue.
Result
Open tool The live tool at playbyplay.football lets users input any game situation and instantly surfaces win-probability estimates for all three options under current overtime rules.
Active projects
Tool
Research
Research
Research
Current work
ToolNFL OT 4th Down Decision Engine
July 30, 2026Football · 2025-26
A decision-support tool that recommends whether NFL coaches should go for it, punt, or kick a field goal on 4th down in overtime — trained on 350,000+ plays from 2016–2024.
PythonXGBoostReact
ResearchA Quantitative Framework for Assessing Wide Receiver Blocking Effectiveness Using Player Tracking Data
January 3, 2026Football · 2025-26
A tracking-data framework for evaluating wide receiver and skill-player downfield blocking, combining defender angles, speed, ball geometry, and blocker-defender movement into a Blocking Effectiveness Score.
Pythonpandasmatplotlib
ResearchQB Pocket Clutch Ratings
May 15, 2026Football · 2025-26
A quarterback clutch metric that compares actual EPA to an XGBoost expectation from pocket geometry, pass-rush spacing, and game context, then isolates how each QB performs in high-leverage fourth-quarter and overtime plays.
PythonnflfastRXGBoost
ResearchNFL Gravity Metric
August 6, 2026Football · 2025-26
A pass-rusher gravity metric that measures observed blocker attention during true pass sets, models expected attention from Big Data Bowl tracking features, and ranks defenders by the extra attention they command.
PythonXGBoostGNN
Archive
Past work
Past work will appear here as the archive is filled out.
Team
Football roster
Eshaan Dhavala
Football Co-Chair
Stats & Data Science · 2027
Abhi Kumar
Football Co-Chair
Math & Economics · 2028
Vardaan Sinha
Football Analyst
Applied Math · 2027
Moulik Chatterjee
Football Analyst
Business Economics; Stats & Data Science · 2027
Mateo Shelton
Football Analyst
Stats & Data Science · 2027
Anika Soitkar
Football Analyst
Data Theory · 2028
Michael Zheng
Football Analyst
Data Theory · 2028
Ayush Verma
Football Analyst
·
Keith Bui
Football Analyst
Data Theory · 2027
Maia Salti
Football Analyst
Applied Math · 2026
Andrew Yang
Football Analyst
Applied Math · 2027