Is there a formula for a hit movie?

Every widely noticed US release, one row each

Did a film make its year's top 40 at the US box office? Each year about 37 films in the list did and about 130 didn't, so every hit is compared with its own year. The hit rate, 22.2%, is almost the same as the song study's 23.1%.

Money and machinery first, then a big, happy story

Logistic regression with a fixed effect for each year and standard errors clustered by lead star. Only features that survive Benjamini–Hochberg correction are listed. "Budget not published" is partly a consequence of success, because hits get their budgets reported, so read it as a caveat, not a lever.

Feature	Odds ratio	95% CI	Unit
Budget (log)	2.47	2.01–3.04	per SD
Sequel / spin-off	2.30	1.82–2.89	yes vs no
US production	2.04	1.60–2.59	yes vs no
Major studio	2.02	1.73–2.35	yes vs no
Genre: animation	1.99	1.30–3.04	yes vs no
Summer release (May–Jul)	1.84	1.52–2.24	yes vs no
Genre: family	1.71	1.14–2.56	yes vs no
Running time	1.69	1.52–1.88	per SD
Happy ending	1.67	1.35–2.07	yes vs no
Spectacle: high	1.66	1.28–2.16	yes vs no
Star's prior hits	1.17	1.09–1.26	per SD
Genre: sci-fi	0.68	0.51–0.89	yes vs no
Audience: adult	0.63	0.48–0.83	yes vs no
Setting: period	0.60	0.47–0.77	yes vs no
Budget not published	0.24	0.18–0.31	yes vs no
Unrated	0.12	0.04–0.40	yes vs no

Picks the bigger hit six times in seven

Every film was scored by a model trained on the other 39 years. Within a year, the full model ranks a real hit above a non-hit 86% of the time. The song model managed 65%. The IMDb rating is hindsight, because audiences rate films after seeing them, so it's reported separately.

Is it just the money?

Budget does most of the work, but the story stands up on its own. Among films with a published budget, budget alone scores 0.80 and the story alone 0.79; together, 0.84. Dropping the leaky "budget not published" flag leaves 0.85. Among studio-scale films only, the full model scores 0.80.

The top-scored 10% of films are hits 79% of the time

Budget matters more every decade; spectacle matters less

Money became destiny

Budget's odds ratio per SD went from 1.4 in the 1980s to 3.47 in the 2000s and 3.03 in the 2010s.

Spectacle stopped standing out

2.54 in the 1980s, 1.11 by the 2010s, once nearly every big film had it.

Sequels never went out of style

2.63 in the 1980s, 2.91 in the 2010s. Romance went from 1.93 in the 1990s to 0.46.

The model's ten favourites: 9 hits and one famous flop

The template is a big-budget summer family sequel from a major studio. The one miss is Heaven's Gate (1981), the $44m western that nearly sank United Artists. The model saw a big, long, prestigious studio epic, just as its backers did.

10 story factors; spectacle pays most

Factor analysis (varimax) on the 46 story codes. Odds ratios are per SD of factor score, within year. As with the songs, the factors predict a little less well than the raw codes (AUC 0.729 vs 0.802). Open a factor to see the films at each end.

Twelve kinds of film; the serious adult drama is the hardest sell

Gaussian-mixture clusters on the factor scores. Lift compares each archetype's hit rate with films from the same years. The archetype audiences love most, serious adult drama, hits at 0.37× its share, even though a higher IMDb rating raises any film's odds (×2.94 per SD, with hindsight). The box office rewards reach, and the rating rewards craft.

Films are far easier to call than songs

Both studies use the same design. Songs are cheap to make and compete on sound; films are expensive, and most of the bet is placed before anyone sees them.

The Choice Engine

Pick a film you love, and the engine ranks every other film by how much someone who loves it would value them. It uses a quadratic consumer-surplus score whose shadow prices come from the hit model, with plain cosine similarity alongside as a baseline. A switch adds the business side (budget, studio, release window, track record), and an era dial favours films from the reference's years. Every card shows a synopsis, a logline and hit-signal chips.