The Library
Papers, models, and ideas worth knowing — each with a strategy takeaway and a link to the source.
7 entries
Coaches historically punted and kicked field goals more often than expected value suggests is optimal.
Why it matters
Risk aversion and career incentives can pull coaches away from the choice that maximizes points and win probability.
Fourth-down choices involve uncertainty, risk tolerance, game state, and institutional pressure — not only an expected-value table.
Why it matters
Two coaches can be acting rationally on the same play if their utility functions, model uncertainty, and time-and-score situations differ.
Top draft picks are often overvalued. Trading up tends to cost more in surplus value than it returns.
Why it matters
If pick prices exceed expected on-field surplus, teams that move down systematically beat teams that move up.
Two-point decisions hinge on expected value, win probability, the post-2015 extra-point environment, and specific score states like down 14.
Why it matters
Going for two is one of the cleanest spots where a clear rule beats gut feel — when used in the right situations.
The 3-cone measures agility, change of direction, and body control. It carries some signal, but it is weak on its own as a predictor of long-term NFL success.
Why it matters
Combine drills can shape draft interest and team workouts, but treating any single number as destiny overweights a narrow test.
Expected points and win probability give every play a value. Turning those values into a clean player-level WAR is harder in football than in baseball.
Why it matters
Football outcomes are deeply shared across teammates, so isolating individual contribution requires careful modeling.
Player-tracking data adds context — spacing, speed, separation, and within-play value — that box-score stats cannot capture.
Why it matters
Many football debates are really about what happened between the snap and the tackle. Tracking data lets analysts watch that interval at scale.