Download the data
The current release contains 427 produced feature screenplay records from 2000 to 2026. Each measure has its own eligible sample. No account or email is needed.
The ZIP includes the expanded analysis CSV, aggregate tables, dictionary, inclusion and exclusion records, checksums and a standalone reproduction script. It contains no screenplay text or private file links.
Data published 4 November 2026. Overview updated 4 November 2026.
View this release and its manifest.
Prefer an email link? Request a copy by email. No newsletter subscription.
What is measured
Counts come from deterministic extraction of screenplay layout. Repeatability does not prove extraction accuracy. Flags and missing observations are excluded separately for each measure.
| Measure | Definition | Eligible n |
|---|---|---|
| PDF pages | Held-PDF page count, including covers and front matter. | 427 |
| Scenes | Recognised screenplay scene headings. | 344 |
| Dialogue lines per action line | Rendered dialogue lines divided by rendered action lines; the column widths differ. | 379 |
| Named locations | Distinct normalised location names found in scene headings. | 344 |
| Detected speaking characters | Detected speaking-character cues, including small roles; not a verified casting breakdown. | 227 |
Genre figures below n=20 are withheld. Empty CSV cells and JSON null mean unavailable, not zero. These are descriptive benchmarks, not measures of quality or success.
Reproduce the findings
- Download and unpack the research ZIP for the release you want to check.
- Read its README and dictionary, including the measurement flags and any historical limitations.
- Run the included
reproduce.mjswith Node.js, following the README. Check the supplied checksums before comparing the reproduced tables.
Use the release identifier in your citation. Earlier releases remain available with their original files and correction notices in the publication history.
Scope and licence
This is a sample of available produced feature scripts, not a random sample of all screenplays or a benchmark for unproduced submissions. Drafts and extraction errors can affect measurements. Read the methodology and limitations before comparing cohorts.
The derived data and figures are licensed CC BY 4.0. Attribute The PreList and identify the release used. Screenplay text and third-party imagery are excluded. Private user scripts never enter this research.
For citations, charts and teaching materials, visit research resources. Research questions or corrections: enquiries@teamprelist.com.