Sources and method
Data and methodology
Every published number is read from a pinned artifact set. This page explains the allowed inputs, publication boundary, and uncertainty labels.
Updated
Public data sources
- nflverse
- Play by play, weekly statistics, snap counts, depth charts, injury reports, rosters, schedules, and the schedule weather fields currently used by the pipeline.
- ESPN public API
- Live game state and freshness checks.
- OpenStreetMap / Nominatim
- Stadium coordinates used by the venue table. OpenStreetMap data is available under the Open Database License, and Nominatim was used to resolve the locations.
- Open-Meteo elevation
- Ground elevation for stadium coordinates from the Copernicus DEM GLO-90 terrain model. This one-time venue-table source is attributed under CC-BY 4.0.
- Kickoff weather status
- Hourly kickoff-window weather is not wired into the pipeline. Current temperature and wind fields come from nflverse schedules, and precipitation remains unavailable.
How published data works
- Pin the edition. A latest pointer selects one complete, published run.
- Read the artifacts. Static pages consume the rankings, games, and track record files from that run.
- Keep provenance visible. Projection cells retain the source and source detail attached by the pipeline.
- Keep calls immutable. Results are scored against the published call after kickoff.
Current source: run.json for season 2026, week 1, run 2026-01-inactives-b2beffbd.
No published benchmark coverage
No public benchmark measurements are included in this published edition.
Uncertainty and calibration
Player surfaces name p20, median, and p80. Game surfaces name the percentile endpoints present in the artifact and pair every likelihood with the research disclaimer. The track record keeps outcomes beside the original call so confidence can be checked over time.
Questions readers ask
What does p20 to p80 mean?
p20 is the lower published percentile, p80 is the upper published percentile, and the median is the center marker. The range states uncertainty rather than hiding it.
Where does the data come from?
The public inputs are attributed on this page. Published measurements and projections are read from the pinned pipeline artifact set.
Is this betting advice?
Sporacle publishes model-estimated probabilities as a calibration research project. Nothing here is betting advice. If you gamble, these numbers are not a reason to.