How the evidence is assembled—and where it stops.

01 · Source and attribution

Public play-by-play, retained as evidence.

DriveLens uses play-by-play from nflverse/nflfastR . The source data is provided under the Creative Commons Attribution 4.0 International license (CC BY 4.0).

Selected penalty totals are cross-checked against ESPN boxscores. That comparison is a practical quality check, not independent certification of the underlying data or every derived result.

02 · Evidence model

Descriptions and record IDs stay attached.

Source play descriptions and record IDs underpin the evidence shown for a result. They let a researcher move from a filtered pattern back to the relevant source record.

A count of fouls is not always a count of distinct affected plays. Multiple fouls can occur on one play, so DriveLens distinguishes those units when the research question requires it.

03 · Comparisons

Matching filters before comparison.

Team and pooled league views use matching filters for the situation under review. This keeps opportunities comparable within the chosen definition.

These comparisons are descriptive. They are not opponent-adjusted models, predictions, or causal estimates.

04 · Documented coverage

Coverage at the documented date.

Ask about available coverage when requesting a pilot.

05 · Interpretation and limits

Unknown stays unknown.

Missing source information remains unknown rather than being inferred. Observed possession outcomes are not causal claims. DriveLens describes what happened and how often; it does not establish that a preceding play caused a later result.

DriveLens does not include game footage or an independently measured time-to-sack dataset.