Data sources & methodology
The sole data source is data/season.json, generated from the public nflverse/nfldata regular-season games.csv dataset. The build validates 32 teams, 272 games and 17 scheduled appearances per team. Dates follow the source’s game-day calendar.
Recorded score pairs from earlier US game-days are used for actual results. This is a daily snapshot, not a live score service. The source has no final-status flag: updates reject scored games on the current US game-day and any disappearance of a previously recorded result. Daily refreshes run outside normal game windows.
Unpicked games are independent 50/50 wins or losses. User-selected ties are supported. No market prices, sportsbook integrations or team-strength forecast are included. Monte Carlo results are estimates with sampling uncertainty.
NFL division and wild-card tiebreakers are implemented separately. If projected scores or touchdowns are required but unavailable, the displayed order is provisional and simulations randomly resolve that unresolved group. The tool reports the number of runs affected.
Clinched and eliminated labels require conservative mathematical bounds or exhaustive remaining win/loss/tie enumeration. Extreme-case simulations alone are not proof. A guaranteed win threshold is sufficient, not necessarily the minimum.
Scenario links store two bits per scheduled game plus a season and schedule fingerprint. Updated actual results override stale selections. New schedule identities invalidate old links rather than silently assigning picks to other games.