On first down, an NFL ball-carrier should stop shy of the first-down marker unless several extra yards are on offer. The decision becomes clear once you calculate the implied yards per possession — a lens that also sheds light on second- and third-down strategy.
🏈 Interactive site → firstdown.microprediction.org
Six interactive figures walk through the argument in expected points: a drive-value
curve with an optimum before the marker, the field-position exchange rate, the
central "what would a possession have to be worth?" calculation (≈ 66 yards to justify
reaching), a 2nd-and-1 vs 1st-and-10 simulation, the 3rd-and-1 run/pass decision, and
the distribution of yards gained. Source for the site is in docs/.
This work appears as a book chapter and a magazine article:
- 📕 Book chapter — Cotton, P. “Stop Shy of the First Down,” in Sports Analytics, L. C. MacLean & W. T. Ziemba (eds.), World Scientific Series in Finance, Vol. 18 (World Scientific, 2021). worldscientific.com/worldscibooks/10.1142/12566
- 📄 Magazine — Cotton, P. “Stop Shy of the First Down,” Wilmott, Jan 2022,
pp. 44–49.
wilmott_paper/ - 📝 Earlier draft (2021) —
Stop_Shy_of_the_First_Down_2021_07.pdf - 🎓 MIT Sloan Sports Analytics submission —
Stop_Shy_MIT_submission.docx - ✍️ Discussion — microprediction.com/blog/nine
| Notebook | What it covers |
|---|---|
Stop_shy_of_the_first_down.ipynb  |
The main analysis |
breakdown_by_play_call.ipynb |
Run/pass breakdowns |
more_comparisons.ipynb |
Additional comparisons |
first_down_images_for_paper.ipynb |
Figures for the paper |
NFL play-by-play for the 2009–2013 seasons from Ben Dilday's
nflMarkov — used here via the
microprediction fork
(inputData/pbp_nfldb_2009_2013.csv), which is built on the
nfldb project.
The chapter sits in Sports Analytics (MacLean & Ziemba, eds.) alongside work on baseball, basketball, hockey, NFL football, and horseracing. If you find this useful, consider following microprediction on LinkedIn.
Photo by Fredrick Lee on Unsplash