EVERY ANGLE,
EVERY ERA.
A century of cinema, indexed across 74,000+ films and seven purpose-built lenses. Open the engine cover. Watch the math run.
What you actually get when you sign in. No marketing inflation.
THREE CHARTS, REAL NUMBERS
Not screenshots. These render straight off the live catalog the moment the page loads. Open any of them full-size in the dashboard.
EVERY ANGLE, A LENS
Each one rebuilt this year as a single, opinionated question. Open any of them straight from the gallery.
FOUR THINGS WE DIDN'T EXPECT
Run the catalog through the right lens and the patterns surface. Each of these turned up during this year's rebuild.
Comedies are 22% less likely to win Best Picture than dramas of equal master score. The Academy has had a tone bias for forty years, and the numbers carry it.
Cinema's average quality score peaked in 1976, not 1939 or 1994. The post-Hays, pre-blockbuster window held the highest mean across critics and audience together.
Repeat director and actor pairings produce films with master scores 1.6× higher on average than first-time collaborations of comparable budget.
When critics and audiences disagree by ten points or more, the audience is right about future box office 71% of the time. Critic consensus undercounts genre.
HOW THE NUMBERS GET MADE
Three layers of analytical machinery, all of them documented. We tell you exactly where the math is sturdy and exactly where it is still maturing.
Every screenplay we touch goes through a quantitative reader that pulls 20 measurable dimensions: scene count, dialogue ratio, vocabulary richness, sentiment arcs, character dominance, and the rest. The features were chosen because they correlate with outcomes, not because they sound impressive.
A Random Forest trained on 4,800 labelled screenplays assigns each new script a quality tier from S through E. The model has a thousand decision trees. Each tree votes. We surface the consensus, the disagreement, and the features that swung the call. Not a black box.
Box-office prediction runs a thousand autonomous audience-persona agents through three rounds of deliberation. The MiroFish engine publishes its forecast every Friday, reflects against actual numbers every Tuesday, and the methodology is in the open. Honest hit rate: 49% overall, 69% on action films.
We say 74,571 films, not 158,000. We say 49% overall hit rate, not 84%. We say 90+ screenwriter style guides, not 429 (that number sums screenwriters, directors, and DPs).
The honest numbers are why studios sign up. Inflated ones blow up on first review. Read the full methodology breakdown for the math behind every claim.
OPEN THE
INSTRUMENT PANEL.
Seven lenses, twenty metrics, a century of cinema. Free to look around.