2 Minute Drill

An NFL “2-minute drill” simulator: you’re down 4 with 2:00 left at your own 25. You draft a 5-man offense (QB, RB, WR1, WR2, TE) from historical players, then face a legendary defense (e.g., 2008 Steelers “Steel Curtain II”) and try to reach the end zone before time runs out.

Inspiration. Late-game football decisions come down to probabilities, and I wanted to make that feeling playable while grounding each play in real career stats.

How it works.

Draft room: Each slot offers 3 randomly chosen historical candidates, shown with career stats (completion %, YPA, catch %, rushing YPC). Simulation engine: Each play draws yardage from a Gaussian distribution via the Box-Muller transform. The mean comes from the player’s career average scaled by the defense’s run or pass rating, so stronger defenses shrink the expected gain. Clock and situation logic: Incomplete passes stop the clock (5s), completions and runs take 12–28s, and under 30 seconds the QB takes deeper shots. With 5 seconds left, the QB attempts a Hail Mary. Commentary: Each play gets a play-by-play description, with an optional Gemini-generated announcer line if an API key is set.

Built with. Next.js 14 (App Router), React 18, TypeScript, Tailwind CSS, Framer Motion, a Next.js API route for draft pools and play simulation, and a Python/Node data pipeline that processes nflverse CSVs into a local JSON file of about 3,800 players (2000–2023).

Challenges. Keeping the odds feeling fair, so that a 40% comp player and a 70% comp player both produce plausible drives. Also, making the data pipeline output light enough to ship without a database.

Accomplishments. A complete, playable draft-to-drive loop deployed on Vercel, with an interactive field, clock, and down-and-distance tracking.

What I learned. Modeling sports outcomes with simple distributions is fast to build, but tuning the parameters is where most of the work goes.