CINEMA. EXAMINED.
The stories the industry tells.
The evidence behind them.
A book series from Hollywood Metrics about how movies actually work. Follow the numbers, question the assumptions, and look closer.
Discover the first volume
Screenplay measurements against the finished film’s reception. Select a measure to see what it means.
Total dialogue words detected by the corrected parser.
Corrected parser · composite quality score only. The rebuild retains 9 original measures and adds 3; it does not validate all 20 original measures.
ONE INDUSTRY.
THREE WAYS IN.
BOOK DATA SNAPSHOT
THE FIRST VOLUME
Nº 001
THE HOLLYWOOD RECORD / VOLUME 001BY HOLLYWOOD METRICS
What 74,571 films and 1,571 screenplays reveal about how movies actually work
An investigation into money, reception, screenplays, and the advice an industry gives itself. Every finding comes with the question that matters: what, exactly, did we measure?
Edition details and purchasing are handled by Hollywood Metrics.
INSIDE THE VOLUME
From the record of a finished film to the words on a screenplay’s page. Open a part to explore its chapters.
Start with the limitsThe number this book was built to find, and what it turned out to be
What is actually in the catalogue, and what the catalogue is not
The denominator under every claim anyone has ever made about film money
The best-known finding in film economics, and the correction that turns it upside down
What a production budget purchases, and the one thing it reliably does not
Genre as a financing instrument rather than a description of content
A claim everyone repeats, false for the entire period they were repeating it
The most seductive artifact in the catalogue, and why the correction makes it worse
Two crowds, and the decade their disagreement changed sign
The belief this catalogue most decisively overturns
Not one line of code in this system deflates a dollar, and nor does anyone else's
The master score, taken apart, exactly
The popularity component is not a rating, it is a headcount
What most films in this catalogue are actually scored on
How many independent measurements twenty features actually contain
A reasonable thing to attempt, and how it was attempted
The grades were never craft judgements, and what that does to the question
The central result, stated plainly and then attacked four ways
The instrument was broken, and here is exactly how
Everything this project published as a positive result
The chapter that turns a broken study into a finding
What the text adds over what the greenlight memo already knows
The director, the writer, the studio and the distributor, measured the same way
What an organisation does when it finds its own bug
The method, stated once, so the eight chapters after it can get on with it
The Shawshank Redemption, measured twice, wrong twice, in opposite directions
The largest failure in the catalogue is not a failure
The other tail, which is forty years older and exactly as misleading
One finding that is completely real, and one number that should never have been printed
CODA, and what an exception to a strong finding is usually made of
Where the two audiences disagree most, and what a famous flop's score is made of
The middle of the catalogue, which is what the catalogue is mostly made of
A second bug, and the one case that is still wrong after the fix
The other corpus, and why it is worth measuring
Act one ends on page thirty. Act one ends on page twenty-five.
Twelve topics, four crafts, and some conspicuous gaps
What the text says about how the text was made
Two hundred words of somebody else's writing, and where they are
The cleaning rules, stated once, so every table can be checked
Eleven decades on one template, so the differences are differences
Every genre counted twice, because there are two defensible ways
Four rankings, each with its gate printed
HOW TO READ THE EVIDENCE
Download the evidenceA result is only as useful as the question, the sample, and the instrument that produced it.
The corrected screenplay measures have weak linear relationships with the composite quality score. This finding concerns these measures and this sample. It does not establish that writing, craft, or nonlinear relationships do not matter.
The original parser missed dialogue in much of the corpus. The corrected analysis uses a smaller set of measures. Switch analyses above to see the distinction; original results are included as a historical record.
Only 15.46% of the catalogue—11,530 films—report both a budget and a worldwide gross. They are not a random sample. Reception, revenue, and profit are different outcomes.
The chart reports Pearson correlations. Each value uses films with both the feature and the chosen outcome available; selecting a measure shows that pair’s sample size. Zero correlation means no linear association, not statistical independence. Zoom changes only the visible scale.
The corrected results come from book/research/refixed_full_corpus.csv; the original results come from pipeline/output/features_with_outcomes.csv, through the book’s fact base. This companion uses the 2026-08-31 snapshot, not a live feed.
Film metadata, ratings and vote counts underlying the statistics in this book are derived from public datasets published by IMDb, The Movie Database (TMDB), and Rotten Tomatoes, and from box office figures published by Box Office Mojo and The Numbers. This product uses the TMDB API but is not endorsed or certified by TMDB. Academy Award records are derived from the Academy of Motion Picture Arts and Sciences awards database. All statistics in this book are the author’s own computations; no source database is reproduced. Screenplay measurements are derived from publicly circulated screenplay texts; no screenplay text is reproduced.
The local catalogue does not record an independent review pass for this volume.
CONTINUE THE INVESTIGATION
The books follow the questions. Hollywood Metrics lets you explore the wider film catalogue and its analytics.
Open the film explorer