How the numbers are made

Every number on Pigskin Nightly comes from public data or from a model we describe here in full. If a figure looks wrong, tell us and we will fix it and say so.

Data sources and refresh rates

• Scores, game clock and live situation: public ESPN feeds, every 30 seconds while games are on, every 10 minutes otherwise.
• Play-by-play, drives and box scores: ESPN game summaries, every 60 seconds during games.
• Standings and seeds: every 30 minutes. Injury report: every 15 minutes, with each status change stored with its time. Rosters and depth charts: every 6 hours.
• Expected points, success rate, CPOE, air yards: the open-source nflverse play-by-play, every 30 minutes on game days and every 3 hours otherwise. Snap counts: nflverse participation data, every 6 hours.

Win probability

Live win probability on game pages, the scores page and the ticker is ESPN’s in-game model, stored play by play. A “swing” is the change in the home team’s win probability from one play to the next; the five biggest swings are the plays that moved the game most.

Expected points added (EPA) and success rate

Expected points measure how many points the offense should score on the current drive, given down, distance, field position, time and score. EPA is the change in expected points caused by a play. Team EPA per play counts only pass and run plays, from the regular season unless a filter says otherwise. EPA allowed is the same number from the defense’s side, so lower is better. A play is a success when its EPA is above zero. CPOE is completion percentage over the completion probability of each throw. Definitions of every metric are in the glossary.

Team strength rating

The rating behind playoff odds and strength of schedule expresses each team in points above or below an average team on a neutral field.

• Half of it is point differential per game, half is net EPA per game (EPA produced on offense minus EPA allowed on defense).
• It is shrunk toward zero by games played: rating × games ÷ (games + 3). After 8 games a team keeps 73% of its raw rating.
• Ratings are centered so the league average is zero. Home field is worth 1.5 points.
• For a single game, the expected home margin is the betting line when one is posted, otherwise home rating minus away rating plus home field. The win chance is the normal distribution of that margin with a standard deviation of 13.5 points.

Playoff odds

We play the rest of the season 10,000 times. In each simulated season every team’s strength is first nudged by a random amount (standard deviation 2.5 points, drawn once per season) to reflect that we do not know its true level, then every remaining game is decided from the win chance above. Seeds follow the league format: four division winners and three wild cards per conference, and only the No. 1 seed gets a bye. Ties in the standings are broken with a simplified order: head-to-head record inside the tied group, then division record (when all tied teams share a division) or conference record, then conference record, then a coin flip. The bracket is played to the Super Bowl with the same win chances. The random seed is fixed, so the same data always gives the same numbers. A team at 100% or 0% across all simulations is shown as “in every simulation” or “out in every simulation”, never as clinched or eliminated: a simulation is not a proof, and our tiebreakers are simplified. “Clinched” and “eliminated” appear only when the NFL makes it official (via ESPN standings: x, y, z and * for clinched, e for eliminated).

Strength of schedule

Remaining strength of schedule is the average strength rating of the opponents a team still has to play; played strength of schedule is the same for the games already played. Both use the current ratings, so they move as teams reveal how good they are.

With / Without

The unit is the whole game. A game counts as “with” a player when the player took at least 25% of the snaps of the player’s unit, and as “without” when the player took no snaps and did not play for another team that week. Games in between (a player hurt after a few snaps) count in neither group. A split appears with at least 2 games with and 1 without, and carries a small-sample badge below 3 games on either side.

Injury load

To rank teams by how hurt they are, out and injured reserve count 1, doubtful 0.6 and questionable 0.3. “Key injuries” are players listed first on the depth chart who are out, doubtful or questionable.

Club desks and articles

Each club desk publishes postgame reviews, scouting reports and weekly numbers built from the same data. Every number in an article must come from the data brief of that game or week. Our standards are in the editorial policy.

Citing Pigskin Nightly

You are welcome to quote our numbers with a link to the page used, for example: Pigskin Nightly, “NFL playoff odds”, https://pigskinnightly.com/playoffs. Each main page has a “Cite this page” line with its update time and a plain-text version for tools and AI assistants (add .md to a player, team or game address). A site map for AI assistants is at /llms.txt, and the full data digest at /llms-full.txt.