About PlayerRiskIQ and how the numbers are made
PlayerRiskIQ is an independent NFL injury research project. It turns the public injury report and play-by-play data into a chance that each player misses his team's next game, and shows who takes on the work if he sits.
Who we are
PlayerRiskIQ is run independently. We are not affiliated with, endorsed by or paid by the NFL, its teams, players, any sportsbook or any fantasy platform, and we sell no advertising placements or paid rankings. A public contact address is not yet published.
Data we use
Injury reports, practice participation, snap counts, play-by-play usage and rosters come from nflverse (CC-BY-4.0); every source is listed on the Data sources page. The site rebuilds on a schedule, about every two hours, and each page shows when it was last updated. Our data can lag breaking news and late scratches; always check the final inactive list about 90 minutes before kickoff.
Methodology
Chance of missing
A statistical model trained on past seasons of injury reports, practice participation and injury history estimates the chance each listed player misses his team's next game. The model is frozen during the season (not retuned week to week) and run on a weekly schedule, before the final game status. When this week's report lists a player Out or Doubtful, the card shows our history for that tag instead (of 3,014 Out listings in 2022–2025, none played), with the earlier model estimate next to it.
Usage impact and role impact
Usage impact is a player's share of his team's targets and carries (pass attempts for a quarterback) over the last 3 team games, times his chance of missing. Role impact weighs snap share, starter status and the drop-off to the next man up, times his chance of missing. Tier cut-offs come from 2017–2021 players. The guide explains every number on a card with a worked example.
Who gains if he sits
Teammate gains compare each teammate's share of the work in past games with and without the listed player. They describe where work went before; they are not a promise of where it goes next.
How we test ourselves
Each estimate is logged before the game and graded afterwards in the public prediction log on the home page. Our main tests are set up before we look at results. One example: in a test covering 1,673 games from 2018 to 2025, adding player availability to a game model did not improve predictions of who wins or by how much, so we do not publish team win or score predictions.
Editorial policy
Articles are written from the same data the cards use. Each article states when its figures were taken, links to the peer-reviewed research it cites, and shows both the date it was first published and the date it was last updated. When a figure changes or we find an error, we correct the article and update that date.