1. Hope (호프) is Na Hong-jin’s big-budget sci-fi thriller that marks his English-language debut and his first film in a decade. It stars a mixed Korean and Hollywood cast, including Alicia Vikander, Michael Fassbender and Hoyeon.

2. Colony (군체) is Yeon Sang-ho’s zombie thriller about survivors trapped in a building during a viral outbreak, notable largely for the return of Jun Ji-hyun to cinema screens after an eleven-year absence.

Six years after Parasite (기생충) became the first non-English-language film to win the Academy Award for Best Picture, Korean cinema’s place at the top table feels established.

This got me thinking about what the ‘Parasite Effect’ was. When a film wins an Oscar, does it have a lasting effect on the perception of that nation’s cinema? Or is it more of an isolated incident with no lasting impact?

So I crunched the numbers on ten years of international winners and their effects.

Is there a halo effect for national cinema?

I took the Best International Feature Oscar winner from 2015 to 2025 and pulled their country’s entire film back catalogue i.e. every film from that nation with a Wikipedia page, minus the winner itself.

I then compared the monthly Wikipedia pageviews in the 12 months following the Oscar ceremony with those in the 12 months preceding it.

To control for global trends (including a rather large pandemic), I divided each country’s change by the average of all other tracked countries in the same window.

This reveals that South Korea is the only winning country that generated a meaningful lift for its national cinema. On Wikipedia, the South Korean back catalogue beat the global average by +24%.

Everyone else came in at or below the peer average.

• Mexico’s back catalogue fell roughly 20% after Roma.

• Chile’s halved after A Fantastic Woman.

• Germany managed a modest +8% after All Quiet on the Western Front (albeit Netflix released that film globally a month into the measurement window, so streaming distribution is doing some of the work the Oscar cannot claim credit for on its own).

It seems that in most cases, an Oscar win does not help a national cinema’s output, but that Parasite was an outlier.

What did the South Korean lift look like?

To show this Parasite Effect in practice, let’s look at the Wikipedia data in more detail.

On the chart below, the orange line is per-film monthly Wikipedia views for 1,301 Korean films (Parasite excluded). The blue line is the comparison group, covering films from France, Japan, Germany, Spain and Italy (10,966 films in total). Both are indexed to 100 at January 2019, so what matters is the shape.

Through 2018 and 2019, Korea tracks its peers almost exactly. From February 2020, they split. Korea rises more sharply and remains 45 to 60% above its pre-Parasite baseline by mid-2020. The peer group gets a bump too (the COVID effect of people locked indoors looking things up on Wikipedis more) but drifts back down.

I have marked Squid Game’s September 2021 release to illustrate that the South Korean lift began a full 18 months earlier. So Squid Game wasn’t the trigger, even if it later reinforced the effect.

Did more people watch South Korean movies after Parasite won the Oscar?

This Wikipedia data is useful for tracking whether people were curious about South Korean cinema, but I also wanted to look at who was watching the movies. So I ran another analysis, this time using almost 10 million audience reviews with submission dates.

The pattern across countries is even more striking in the review data. South Korea is again alone at +52%. Iran shows a positive number too, but that is based on just too few reviews in the pre-win window to be a reliable signal.

Every other country with a robust sample came in flat or negative. Mexico and Chile both fell by roughly half. Denmark, Japan and Hungary all ended up well below their peers.

Was it South Korea or was it Bong?

When Parasite won, was global curiosity pulled toward Korean cinema as a whole, or specifically toward Bong Joon-ho? Hulu bundled Mother, The Host and Barking Dogs Never Bite alongside Parasite in April 2020, which gave an obvious reason for a director-specific spike.

The data says national, not personal. Bong’s five pre-Parasite films moved +43% on Wikipedia in the 12 months after the Oscar. The broader South Korean catalogue moved +46%. On user reviews, Bong’s back catalogue barely shifted (+8%) while the wider South Korean catalogue almost doubled (+86%).

So it’s fair to say that this film’s Oscar win contributed in a big way to reshaping how audiences thought about an entire national cinema.

So why South Korea and not the others?

This kind of data can reveal the pattern but not the reasons behind it. Here are my first thoughts as to why this moment played out as it did:

1. The movie is outstanding. No, really. It’s a cut above most films, even at the Oscars that year, and watching art that good can be exhilarating in a way that makes you want to chase that high. The film was infused with a number of South Korean elements, characters, and tropes, which made viewers keen to learn more.

2. The speech. When Bong picked up his Golden Globes, he said something widely reported: “Once you overcome the one-inch tall barrier of subtitles, you will be introduced to so many more amazing films”. It made the whole thing feel more like a moment in history than just a single film winning a vote.

3. Streaming was ready. Neon’s deal with Hulu put three Bong films on a major US platform in April 2020, exactly when lockdowns sent viewers looking for something new. By February 2021, Netflix had committed $500 million to Korean content in a single year, on top of $700 million already spent between 2015 and 2020. That commitment later grew to $2.5 billion through 2027.

4. Korea had a deeper bench. Iran, Chile, Denmark and Hungary do not have Memories of Murder, Train to Busan, Oldboy and The Handmaiden queued up and ready to reward a curious viewer. When Parasite opened the door, there was somewhere for the audience to go.

5. The timing was extraordinary. Parasite won on 9 February 2020, which was roughly five weeks before the world went into a global lockdown. People confined at home with streaming subscriptions and an appetite for novelty are the ideal environment for a back catalogue to catch fire.

Epilogue

For most filmmakers, reaching the heights of winning the Best International Feature at the Oscars is a personal honour. But once in a while, it is also a moment when you get to help your peers and increase a lasting interest in your nation’s cinematic output.

Parasite was a rare case where not only was the film exceptional, but the collision of a compelling argument, a ready infrastructure, a deep catalogue worth discovering, and a world that was about to be locked indoors with nothing to do but watch.

Notes

I tested every Best International Feature winner from 2016 to 2023 (eight films, eight countries). For each, I pulled the winning country’s entire film back catalogue from metadata databases, excluded the winner itself, and compared Wikipedia pageviews and written review volumes from several major review sites in the 12 months after the Oscar ceremony with the 12 months before.

To strip out global trends (such as the odd pandemic) I measured each country’s change relative to the average of all other tracked countries in the same window. The comparison group for South Korea is France, Japan, Germany, Spain and Italy, which I chose for coverage density. The 2015 winner (Ida, Poland) was excluded for insufficient pre-win data, and the 2025 winner (I’m Still Here, Brazil) has not yet completed its post-win window.

Both Wikipedia and the review data skew toward English-speaking audiences, so this measures global cinephile curiosity as expressed in English, not South Korean domestic viewership.

Co-productions can appear in more than one country’s catalogue, which inflates the numbers slightly for Germany and France but is a minor issue for South Korea.

Ted suggested I synthesise them into a series of meta-lessons that recur across the films.

You can read all of Ted’s collected case studies here Distribution Case Study Masterlist v1.0

The case studies span four decades, from Spike Lee’s She’s Gotta Have It in 1986 to Union in 2024. Budgets range from $2,000 to $2.5 million. The films include micro-budget narrative features, impact documentaries, festival darlings, and a few that no distributor would touch.

What follows are the nine patterns I kept seeing recurring in the firsthand accounts of filmmakers who went through it.

1. Build your audience before you need them

The single strongest predictor of distribution success across these case studies was whether the filmmakers had built a relationship with their audience before the release.

Sharon ‘Rocky’ Roggio made 1946: The Mistranslation That Shifted Culture, a documentary about a Bible translation that shifted attitudes towards homosexuality. Netflix passed, but rather than scramble for another platform, she built her own distribution model from scratch, starting while the film was still in production. She identified her core audiences (faith communities, LGBTQIA+ allies, scholars), ran Zoom focus groups, sent regular newsletter updates, and converted super-fans into volunteer ambassadors.

Don’t just make it available — make it matter.
Sharon ‘Rocky’ Roggio, director of 1946 (via Jon Reiss’s case study)

They had over 500 watch parties across 25 countries, net virtual screening revenue of $118,000, a TikTok following of 201,000+ and the whole time she retained all rights and viewer data.

The Wisdom of Trauma directors Zaya and Maurizio Benazzo distributed the film exclusively through their own website, using Vimeo. They grew their email list from 60,000 to 900,000 subscribers during the campaign. Four million views in the first seven days. First-week revenues exceeded the entire film budget and fully financed their next project. Crucially, 50–60% of views came from word-of-mouth, not paid advertising.

Sandi DuBowski, director of Sabbath Queen, has been collecting email addresses at screenings with physical clipboards since releasing Trembling Before G-d in 2001. Twenty-four years of audience data, one clipboard at a time.

Cryptopia director Torsten Hoffmann, whose Kickstarter campaign reached its goal in nine days, found that 70% of funds raised came via email. His was:

The modern filmmaker is more of an entrepreneur who has to create a “product” for a “market”. Audience building and branding are part of this process.
Torsten Hoffmann, director of Cryptopia (via The Film Collaborative case study)

There were also some examples of filmmakers who didn’t do this and later regretted it. Daresha Kyi, director of Mama Bears, reflected that she ‘should have collected audience emails more systematically’.

Julia Kots of Inez & Doug & Kira found that without a pre-existing audience or clear niche, options narrowed fast.

In short, if you haven’t started building your audience by the time you lock picture, you’re already behind.

2. Reject inadequate deals - self-distribution can outperform

Across these case studies, filmmakers who rejected disappointing traditional distribution offers and chose to self-distribute frequently outperformed what they would have earned through conventional deals.

Hundreds of Beavers is the standout. Traditional distributors offered maximum guarantees of $30,000 for a film that cost $150,000 to make. The filmmakers refused and pursued what they called ‘hybrid self-distribution’, splitting rights across theatrical, TVOD, international, and physical media.

Hundreds of Beavers:

• Budget $150,000

• Total revenue ~$800,000 (as of October 2024).

• Distribution cost ~$135,000

• Plus an estimated $150,000–$250,000 of unpaid social media labour.

Had they accepted the $30,000 offer, they would have earned roughly 4% of what they ultimately made.

Union premiered at Sundance, won a jury award, appeared on multiple ‘Top 10 Films of the Year’ lists, and screened in 23 countries. Yet every major distributor passed, and the producers chose to self-distribute.

Distribution is part of making a film and needs to be just as strategic, creative, and contextual to your film as its team, artistic vision, and style.
Samantha Curley, producer of Union (via Distribution Advocates case study)

Their Black Friday TVOD launch through GATHR (a five-day window with the tagline ‘This Black Friday, don’t order from Amazon. Watch UNION’) generated 2,000 rentals and netted as much revenue as the ‘entire theatrical run’ with, as they put it, ‘way less money, time, and resources’.

Burn provides the founding text for this model. After winning the Audience Award at Tribeca in 2012, the best offer was ‘low-ball offers, in the mid five-figure range’. Co-director Brenna Sanchez’s response:

Don’t believe someone else can do it better than you can. And certainly don’t wait for them to do it. Be nimble.
Brenna Sanchez, co-director of Burn (via Film Independent case study)

They four-walled theatres across America. The 25-city northeast tour alone grossed $375,000 and eventually played in 170+ cities. Tom Putnam’s advice:

‘Plan on releasing the movie yourself as part of budgeting; look at distributor pickup as happy accident if deal makes sense.’

Columbus faced the same dilemma. Despite Kogonada’s feature debut receiving ‘exemplary reviews’ at Sundance with John Cho and Haley Lu Richardson, the best all-rights offer was $150,000 against a $700,000 budget. A $550,000 shortfall. They chose creative distribution instead.

None of this is to say that self-distribution always wins. But the default assumption (i.e. that a traditional distributor will do better) is, based on these case studies, increasingly unreliable. Particularly for documentaries and films without a recognisable cast.

3. Eventize every screening

If there’s one tactical recommendation that appears more consistently than any other across these case studies, it’s this: don’t just show your film. Make every screening an event.

Yes, I hate the word ‘eventize’ too. But the concept behind it is the closest thing these case studies have to a universal principle.

Sabbath Queen’s Sandi DuBowski pursued a year-long eventized theatrical model. Nearly every screening was transformed into what he described as a package:

‘It was like a package experience. Oh, the Sabbath Queen Cruise is coming to town — you see the film, then you have a ritual feast, then a deeper discussion.’
Sandi DuBowski, director of Sabbath Queen (via Jon Reiss’s case study)

His screenings included partnerships with houses of worship, restaurant-hosted Shabbat dinners, and ‘Soul Spa’ events. Box office: over $110,000.

Hundreds of Beavers took a different approach: mascots, wrestling, merchandise, and a fur red carpet at the LA premiere, which generated coverage in Variety. They staged a 14-city Midwest road show with Q&As. The events were fun but also provided footage that could be repurposed as paid advertising content.

Eno, Gary Hustwit’s generative documentary about Brian Eno, pushed the concept furthest. Custom software creates a unique version of the film for every screening, so each showing is genuinely unrepeatable. Premium event screenings (where Hustwit performed the film ‘live’ on stage) commanded $25-$40 tickets, with 100% of the proceeds retained by the filmmaker.

There is a reporting issue here that I found particularly interesting (I know that, as a data nerd, numbers will always jump out at me, but in this case, I think it’s worth noting for everyone). Eno’s total theatrical gross was over $1 million. But the standard box-office reporting showed only $298,989. The remaining $400,000+ came from event screenings and partnerships that trades don’t count. The gap between reported and actual revenue was enormous, and invisible to anyone reading the trade press.

It makes you wonder what else we’re missing.

The Twin Seas Media case studies demonstrate this model at scale for impact documentaries. In the Whale reached 10,000+ viewers across 72 screening partners in 13 states over 12 months, with ‘sold-out screenings and encore requests’ described as ‘frequent occurrences’. A Reckoning in Boston reached 10,000+ people through 65+ partners across 28 states, generating $100,000 in licensing and speaking fees. Same model every time: screening + Q&A + community conversation + local partner.

Even at the micro level, Guacamole Yesterdays organised a multi-city tour pairing screenings with Q&A sessions and community events. They eventually secured placement for Apple TV pre-orders while retaining control over pricing and audience engagement.

Overall, it seemed to me that a screening people attend generates more revenue, more word-of-mouth, and more downstream value than a screening they merely watch.

4. Constraints are creative fuel, not limitations

We should acknowledge at this point that with any list of micro-budge case studies, there is survivorship bias. By their very nature these are the success stories, and for every Clerks there are hundreds of micro-budget films nobody ever heard of.

The case studies are only of films that beat the odds. But what’s interesting is how they beat the odds. In almost every case, the budget constraints produced a visual identity that audiences celebrated rather than forgave.

Read more

It’s a question I’ve looked at before on this site, most recently in 2019, but a lot has happened in the past six years. One pandemic, two strikes, three Spider-Men, a four-hour Justice League and a flood of streaming.

A recent poll of 2,000 Americans found the ‘ideal’ movie length is 92 minutes. Only 2% said they were comfortable with films over two and a half hours. And yet nine of the ten highest-grossing films of all time run over two hours, and three of them cross the three-hour mark.

So what’s going on? Let’s look at the numbers…

The headline paradox

If you look at every film produced worldwide, the average running time has barely changed in decades. It has hovered around 100 to 103 minutes since the 1980s. In 2024, the average film was 103.6 minutes long.

If this doesn’t match your experiences, that’s because this is all the films released, not the ones people actually watch.

Wide theatrical releases (the films that open in hundreds or thousands of cinemas across North America) averaged 106 minutes in the early 2000s. By the 2020s, that figure had risen to 114 minutes.

This means that if you’re a regular cinemagoer, you are now routinely sitting through films that are around ten minutes longer than they were a generation ago.

And the biggest-budget films ($100M+) have pushed even further, pulling the theatrical average up with them.

By the way, pre-show trailers and adverts have ballooned too. Most major chains now run 20 to 30 minutes of material before the feature starts, meaning that your cinema trip has got longer at both ends.

The two-hour film is taking over

Averages can mask what’s really going on underneath. So I split every wide release into runtime buckets and compared the decades.

In the 1980s, around 14% of wide releases ran over two hours. In the 2020s, it’s 32%.

At the other end, the ‘short feature film’ (i.e. a commercial feature film, rather than a “Short Film” which students and early stage filmmakers might make for practice, showcase or ego) has all but vanished from mainstream cinema.

In the 1980s, roughly 13% of wide releases ran under 90 minutes. In the 2020s, that’s down to 7%.

Initially, I thought this might be short films sneaking into the data, but no, it seems shorter feature films are dying out. The vast majority are 75 to 89-minute features. Animated family films like The Land Before Time at 69 minutes, the odd Woody Allen at 79, that sort of thing.

The whole distribution has shifted upwards. While there are some outliers dragging the average up, it’s more the case that the entire middle ground has moved. The 105-to-119-minute film is now the most common runtime bracket for a wide release, and the two-hour-plus film is no longer unusual.

Very long films have proliferated dramatically. In the entire 2010s, only a handful of mainstream Hollywood films crossed the three-hour mark.

In the past couple of years, we’ve had The Batman (175 min), Oppenheimer (180 min), Avatar: The Way of Water (190 min), Killers of the Flower Moon (206 min) and The Brutalist (215 min).

Which genres are responsible?

Not all genres have inflated equally. Action films have grown the most (hence why Matt invited Todd to speak about it on the show).

The average wide-release action film ran 103 minutes in the 1980s. By the 2020s, that had increased to 128 minutes, up 25 minutes.

Comedy and horror have crept up too, but not to the same degree. Wide-release comedies held steady at 99 minutes from the 1980s through the 2000s, then nudged up to 107 minutes in the 2020s. Horror went from 97 minutes to 103. C

Overall, we can see that runtime inflation is concentrated in genres with the highest budgets and set-piece costs.

This makes sense. A comedy needs to land its jokes and get out, and a horror film needs to sustain tension, not test endurance (like Get Out). When I last looked at horror runtimes in 2019, It: Chapter Two at 169 minutes was longer than 99.78% of all horror films made in the prior 20 years.

Even animation has crept up. Disney Renaissance films in the 1990s typically ran 83 to 95 minutes (The Lion King is 88 minutes, Aladdin 90, Beauty and the Beast 84). Modern Disney and Pixar features cluster around 100 to 118 (Moana 113, The Incredibles 115, Incredibles 2 118).

The franchise escalation effect

The obvious suspect is the franchise machine. Surely it’s all those Marvel sequels and cinematic universes bloating the schedule?

Kinda. In the 1980s and 1990s, franchise films were actually shorter than standalones by about three minutes. The typical franchise entry was a tight, efficient sequel like Lethal Weapon 2 or Home Alone 2.

That changed during the 2000s and has been widening since. By the 2020s, franchise films averaged 121 minutes versus 112 for standalones (i.e. a nine-minute gap).

The crucial point is that both have got longer, with standalones going from 105 minutes to 112 minutes, and franchise films from 102 to 121.

What the franchise data does reveal is a pattern of escalation within individual series. Each instalment tends to run longer than the last, as though runtime is part of the promise that each sequel will be bigger than its predecessor.

• The Star Wars sequel trilogy averaged 143 minutes per film, compared with 125 minutes for the original trilogy (18 minutes more per film).

• Indiana Jones and the Dial of Destiny (142 min) ran 27 minutes longer than Raiders of the Lost Ark (115 min).

• No Time to Die (163 min) was the longest Bond film in the franchise’s 60-year history, more than 50 minutes longer than Dr. No.

• The Fast & Furious franchise hasn’t had a sub-two-hour entry since 2009. Every film since Fast Five has run 130 minutes or more.

Follow the money

Budget turns out to be a stronger predictor of runtime than franchise status, genre, or almost anything else.

Films made for under $10 million have barely changed in length over four decades. But films budgeted at $100 million or more now average about 130 minutes, up from around 122 minutes in the 2000s.

My read on this is that a studio spending $200 million on a film wants every set piece on screen and for the audience to feel they got their money’s worth from a premium-priced cinema ticket. Cutting twenty minutes might make for a tighter film, but it doesn’t make for a bigger event.

There's a strange quirk in how film exhibition works. In most cinemas worldwide, the ticket price is unrelated to the runtime. A three-hour film consumes 50% more screen time than a two-hour film, but the ticket costs the same.

There are exceptions:

• Germany has a formal surcharge system (Überlänge) used by major chains, with fees kicking in at 120 minutes and rising in brackets.

• Kinepolis in Belgium adds €1 for films over 130 minutes.

• Viewshow Cinemas, Taiwan's largest chain, charges a tiered premium starting at NT$20 (about a dollar) for films over 150 minutes.

But these are outliers. The global norm is flat pricing, regardless of runtime. A three-hour film consumes 50% more screen time than a two-hour film, but tickets tend to cost the same.

From the audience’s perspective, a longer film is arguably a better value. From the exhibitor’s perspective, it’s a structural subsidy for long films that nobody has ever corrected.

So what’s going on?

I don’t have a single neat answer, but there are a number of things worth pointing to.

1. Digital projection removed the practical ceiling. Until 2017, most cinemas projected from 35mm film reels. Now they don’t, and a digital file can be any runtime. A standard reel held roughly 11 minutes of footage, so a movie like Avengers: Endgame at 181 minutes would have required 16 reel changes.

2. The economics of theatrical cinema have changed. Cinemas need each screening to justify a premium ticket price, and maybe a longer film can feel like a better value.

3. Studios want their big releases to feel like events. A three-hour epic signals importance in a way that a tidy 95-minute film doesn’t.

4. Shorter films have migrated to streaming. Between 1995 and 2009, Hollywood’s major studios averaged roughly 112 theatrical releases per year. Between 2010 and 2023, that fell to about 83.

5. The dual-window model has given directors a release valve. Ridley Scott delivered a 157-minute theatrical cut of Napoleon for Sony, followed by a 205-minute director’s cut for Apple TV+. The theatrical version plays the multiplexes while the extended version lives on streaming.

6. Directors with leverage are using it. Christopher Nolan, Denis Villeneuve, Martin Scorsese. These are filmmakers with the clout to deliver a three-hour cut and have the studio release it. In an earlier era, the studio might have insisted on cuts. Today, a long runtime from a prestige director is a selling point, not a problem.

7. And audiences keep showing up. Oppenheimer ran three hours and grossed $952 million worldwide. Avengers: Endgame ran 181 minutes and crossed $2.7 billion. Avatar: The Way of Water ran 190 minutes and made $2.3 billion. If audiences were punishing long films at the box office, studios would have course-corrected by now.

I looked at that penultimate point a few years ago in an article titled "Do more experienced directors make longer movies?" In a word, yes!

Epilogue - The intermission question

If films are going to keep getting longer, should cinemas bring back the intermission?

Intermissions were standard in cinema for decades. Gone with the Wind, Lawrence of Arabia, Ben-Hur, The Sound of Music all had built-in breaks. The last major wide release to include one was Gandhi in 1982. After that, they disappeared from Western cinemas entirely because exhibitors calculated that dropping the break would squeeze in one more screening per day.

Then in 2024, Brady Corbet’s The Brutalist brought it back. At 215 minutes, the film includes a 15-minute intermission designed into the script at a narrative pivot point. The reception was overwhelmingly positive and sparked a conversation about whether more films should follow.

Not everyone agrees. When an independent cinema in Colorado added an eight-minute break to screenings of Killers of the Flower Moon (206 minutes), Paramount and Apple contacted them and told them to stop. Scorsese reportedly made it clear that an imposed intermission was an insult to the film.

In India, of course, intermissions never went away. Every Hollywood film released in Indian cinemas gets a mandatory interval, including 84-minute animated films. Concession revenue during intervals accounts for 30 to 40% of Indian cinema revenue. Christopher Nolan has pushed for his films to be shown without breaks, but many Indian chains simply ignore him.

It’s a tension that isn’t going to resolve itself. Directors see the intermission as a violation of their artistic intent. Audiences (and their bladders) increasingly need one. As films push past three hours more and more regularly, the pressure will only grow.

Notes

The raw data for today’s research came from the OpusData/The Numbers database, which covers 36,431 films with runtime data. ‘Wide releases’ are defined as films with a wide release pattern in North America, as classified by The Numbers. Budget data is available for roughly 30-40% of theatrical releases and coverage drops to around 20% for the 2020s, so budget-based findings should be read with that caveat in mind.

For most of movie history, a movie had one producer.

In the 1940s and 1950s, the vast majority of movies credited just one producer. Even through the 1970s, producer teams were small and stable.

That world is gone.

In 1940, almost three-quarters of films were produced by a single producer, whereas by 2025, that had fallen to barely a quarter.

Today, films average 4.5 producers per credited film, up from 3.0 in 2000. This is a 50% increase in just over two decades.

Not all producers' credits are created equally. There are six major types of producers:

1. Producer is the (or one of the) overall project lead(s), akin to a CEO.

2. Executive Producers operate at a high level, not day-to-day work.

3. Co-Producers help get the film made, likely with a single focus.

4. Line Producers are all about the mechanics of the film shoot.

5. Assistant Producers are the most junior of the producing credits involved with the making of the film.

6. Associate Producer is a nebulous term of diminishing value.

I’ll go through each in a little more detail as we come to their data.

The “Producer” credit

When someone gets a “Producer” credit, they are rightly seen as a key creative and managerial lead on the project.

Regardless of whether it's one or multiple people receiving this credit for a movie, it signals that they were key to making the film. This will most likely include developing the project, assembling the key team, securing financing, and overseeing the film from development through release.

The “Executive Producer” credit

An Executive Producer is traditionally a senior figure who helps secure financing, studio backing, rights, or the attachment of major talent to the film. Their contribution is most likely to be at a high level rather than in day-to-day production management.

The number of Executive Producer credits has more than doubled since 2000, up 111.9% over the past century.

Other producing credits

Three of the remaining major producing credit types are more directly tied to the making of the movie.

• Co-Producers are supporting producing partners who share responsibility for delivering the film, often managing a specific part of production or representing a collaborating company or production entity. So if you film in France for some of the scenes, then you might have a French Co-producer who was the lead on that part of the shoot (or financing).

• Line Producers are responsible for the practical execution of the film, including budgeting, scheduling, hiring crew, and ensuring the production runs on time and on budget.

• Assistant Producer credits are an entry-level producing credit, often given to someone stepping up into producing for the first time or taking a big leap in their career. It will be assumed they worked hard on making the movie.

That leaves us with the Associate Producer credit. I understand that in the middle of the last century, this credit often did the job we now use “Co-Producer,” i.e., a partner for a particular aspect of the production.

However, in modern parlance, it’s all too often regarded as a variety credit, given to placate someone who needed to be brought onside (e.g., access to money, talent, or location). In David Mamet’s brilliant film about filmmaking, State and Main, they describe the Associate Producer credit as being “what you give your secretary instead of a raise”.

Interestingly, Line Producer credits have not changed much in recent years (a 16.8% increase between 2000 and 2025). This is likely because filmmaking has not changed in ways that would require additional Line Producers, but also because they’re not seen as attractive credits by those looking for a title bump.

Not so the other three.

• Co-Producer credits increased by 36.9%

• Associate Producer credits are up by 59.9%

• Assistant Producers have almost doubled, up 80.8%.

How the scale of production affects the number of producers

Generally speaking, films grow in complexity as their budgets increase. It’s not linear, nor universal, but a film costing over $100 million will have more moving parts than a micro-budget production. Therefore, it is perhaps not surprising that the number of producers increases with budget level.

Films costing over $50 million have, over the past quarter century, averaged 10.2 producers (of all types).

The greatest increase is among Executive Producers as we move up the budget levels. On average, a film costing over $50 million will have more than twice as many Execs as a film costing under $20 million.

How do you get to be a producer?

In the simplest terms, you produce a movie, and you slap your name on it. That means you:

• Find a script (i.e. write one, buy one, or hire a writer)

• Raise the money from investors, tax credits, and loan sharks.

• Spend the money hiring people to plan, shoot, edit and finish the film.

• Hopefully, do deals to release the film in festivals, cinemas and home entertainment.

• You submit your credit to IMDb. Voilà - you are now a producer. (Technically, you were a producer the whole time, Dorothy, as the act of producing is what makes you a producer, but this is the point at which you get the show off at parties).

• ???

• Profit

However, in the real world, things get a lot more complex. On most films, a vast number of people and companies work together to make a film happen. Who exactly has the right to anoint someone with a producing credit will depend on the exact contractual relationship between the parties.

Receiving a producing credit is one of the levers that people have to close a deal. I.e., you might say to an investor that they will get an “Executive Producer” credit if they invest money in your film.

And in other cases, credits are awarded in exchange for securing scripts, IP, talent, money, distribution deals, locations, and more. Anytime a real producer wanted to sweeten a deal, they could offer some form of producing credit.

Which meant that people started demanding it, seeing it as a shortcut to status, respect, and showing off at their high school reunion.

Acting like a producer

One aspect of this inflation is the growth of major actors also receiving producing credits. There are a number of notable actors who are also prolific producers in their own right. Few members of the public realise that some of their favourite stars, such as Danny DeVito, Robert Redford, or Clint Eastwood, are also respected producers.

Even if they did not start out producing movies, sometimes an actor can gain so much power that they rise above their ‘acting’ label to become the main driver of a franchise. Rather than saying “Tom Cruise acts in Mission Impossible”, it’s fairer to say that “Tom Cruise IS Mission Impossible”. His work behind the scenes of the franchise is producing at its purest.

In the 1940s and 50s, it was rare for an actor to be credited as a producer of any type. The rate of actor-producers rose sharply in the 1960s and 1990s, nearly doubling between 1986 and 2009.

Over the past two years, we’ve seen this level off to the point that about one in eight people with a producing credit on a movie also receive an acting credit.

As well as the rise of genuine hyphenates, part of this huge growth has been the trend of major actors demanding more status and attention by also securing producing credits.

Amanda Seyfried recently admitted that she had no idea she was an executive producer on The Housemaid until she spotted it on the call sheet three weeks into filming, saying that her agent had negotiated the credit without telling her. She said:

It’s one of those vanity credits… I didn’t do shit to make that movie. I only acted in it.

To be fair, I give Seyfried credit for admitting publicly that it was an empty title (great, now I’ve added to the problem by giving her another credit), although she could have done what Scarlett Johansson did and asked to have her Exec Producer credit removed when she didn’t think she earned it on Thunderbolts*.

As credits were given away freely, inflation took hold

This fragmented, power-based, bespoke credit system is why we’re seeing such credit inflation. Put simply, no one is regulating it. Anyone can make anyone else “a producer” at any time for any reason.

This is very different to most other major roles in Hollywood. The Writers Guild of America (WGA) determines writing credits, and the Directors Guild of America (DGA) determines directing credits through formal arbitration systems, but producers are typically classified as management rather than labour. This historically left the assignment of producer credits largely in the hands of studios and financiers.

The writers’ and directors’ crediting rules arose because creators felt that producers were abusing their positions to give themselves credit. (This is a time in which arrogant producers stole writing credits for themselves, while the humble producers stole writing credits under a pseudonym). However, with producers, there is no third party - they’re doing it among themselves.

On a film-to-film basis, this wasn’t a problem, but as credit inflation took hold, it was devaluing the whole idea of a credit. This came to a head at the Oscars in the late 1990s.

The first “Academy Award for Best Picture” were given to a film’s production company, but starting in 1950, the rules changed so that it went to the individual producers. Anyone with a “Produced by” credit was considered a winner and received a statuette.

When Shakespeare in Love won in 1998, five producers collected the award, even though later accounts said only two of them (David Parfitt and Donna Gigliotti) had done the traditional producing work. Of the other three:

• Harvey Weinstein had taken a producer credit despite limited on-set involvement.

• Marc Norman’s producer credit had been negotiated by his agent.

• Edward Zwick retained his producing credit despite leaving the film.

In response, the Academy introduced a new rule limiting the number of producers eligible for the Best Picture award to three. In order to be counted as an Oscar winner, producers now need to be one of up to three people who have an on-screen credit of “Producer” or “Produced by”, and have “performed the major portion of the producing functions”. If the producers of a movie can't agree among themselves who the three chosen should be, then the Academy’s Producers Branch steps in to make a ruling.

The Academy is not the only Hollywood body trying to determine the difference between real and variety-producing credits. The Producers Guild of America (PGA) introduced the “p.g.a.” Producers Mark in 2012, certifying which credited producers actually performed the core producing functions. Despite this, the “p.g.a.” mark is voluntary and does not prevent studios from adding additional producer credits.

Both bodies are trying to fight the tide from their little beacheads. In both cases, the rules are clear, defined, and fair. Yet producing a movie is messy and busy, and there will always be fringe cases, including:

• Bob Yari assembled the financing for Crash and received a full on-screen “Producer” credit, but when he was excluded from Oscar recognition, he launched a series of lawsuits against the Producers Guild and the Academy. He argued that their opaque process had reduced him to a mere “money man” and cost him both prestige and financial upside. He lost on appeal.

• Long‑time producing partners Ron Yerxa and Albert Berger developed Little Miss Sunshine from an early script, championed the writer, and helped pick the film’s first‑time feature directors. Five producers were certified by the PGA, but the Academy picked the other three producers as the only official nominees. The pair were gracious at the time, but have since been more open about their frustration in feeling like the Academy’s arbitrary cap was erasing years of their work. The backlash led the Academy to soften its rules to allow more than three producers in “rare and extraordinary” cases.

• Boyhood took 12 years to make, and over that time, Jonathan Sehring and John Sloss provided long‑term financing, logistical support, and protection so that Linklater could shoot for a few weeks each year and retain creative freedom. They both received full “Producer” credits on screen, but the PGA ruled that only Richard Linklater and on‑set producer Cathleen Sutherland qualified for the “produced by” mark.

Why are there more Executive Producers today?

While these efforts to regulate the full “Producer” credit are imperfect, they are at least trying. Until very recently, there has been no comparable movement to regulate Executive Producer credits.

And because of this, the most acute form of this credit inflation is among Execs. As noted at the start of this article, the average number of Executive Producers on movies has increased by 112% since 2000.

This is down to a number of factors, including:

1. Indie financing comes from more sources. Independent films often assemble budgets from multiple investors or financing partners. Each party expects recognition, and Executive Producer is the credit most commonly used.

2. Studio movies are assembled as packages. Modern Hollywood projects are built by attaching talent, rights, and financing before production begins. People who help secure one piece of that package are often rewarded with an Executive Producer credit.

3. Producer credits are unregulated. As we’ve learned, there is no check on studios awarding credits to anyone for any reason. Things like the “p.g.a.” mark and the Oscar limitation add some friction to the wording of “Producer” credits, but not Executive Producer credits.

4. Actor contracts increasingly include Exec Producer titles. Agents have realised it’s something they can demand on behalf of their client, and in many cases, it’s an easier win than actually getting more money. Amanda Seyfried put it best when she said, “I didn’t do shit to make that movie. I only acted in it“.

5. They’re free. Exec Producer credits are used to compensate people without increasing salaries. The industry is in a major state of fear and flux, so giving away credits in lieu of money feels like an obvious solution.

So what now?

With credit inflation becoming rampant, industry efforts to push back are hardening.

In 2024, a group called Producers United launched with around 80 established producers (currently 281) and has since pushed for clearer standards on who actually qualifies for a producing credit.

They want studios to tie “Producer” credits to specific duties, such as participating in prep, sharing budget responsibility, and approving key hires, and to stop using Executive Producer titles as negotiation currency.

Variety neatly summed up their fight with the headline “Hollywood Career Producers Are Battling Actors, Greedy Managers, Penny-Pinching Execs and Stolen Credits to Survive“.

Time will tell how effective they will be. The history of Hollywood is one of people having to fight tooth and nail to protect their rights, credit and value. Nothing is given for free.

What’s new here is that typically, the producers were regarded as the people you need to fight, not the victims.

Notes

It’s worth noting that all of this applies only to movies. Television production has a whole different relationship with producer credits, and especially Executive Producer credits. TV is more of a writer-led industry, so many forms of producer credit relate to writing and showrunning (i.e., managing overall production and the other writers).

The 2025 Argentine film Death of a Comedian cited 10,123 Associate Producers. I excluded this title from the statistics as I thought it was heavily skewing the real picture. I don’t like manually excluding titles, partly for data-science reasons and partly because film is a business of outliers. However, here it swamped the overall picture, and I thought it did a disservice to the reporting of reality to include it.

Despite removing it from the average calculations, I wanted to highlight it here because it is a good example of another way people use producer credits. Thousands of fans purchased small investment bonds that helped fund the film, and each investor received an Associate Producer credit in return. The model was organised through the Comunidad Orsai collective and raised roughly $1.85 million from more than 10,000 micro-investors.

The sheer scale of the credit list created an unexpected technical challenge. IMDb normally requires credits to be entered individually, so manually uploading over ten thousand names would have taken months. Instead, a community programmer wrote an automated script that bulk-submitted the names to IMDb’s system, allowing all investors to appear in the film’s official credits database.

Industry rule = One page of screenplay equals one minute of screen time.

It’s one of those pieces of received wisdom that gets repeated so often it starts to feel like a law of physics. Screenwriting books state it as fact. Coverage readers rely on it. Studio executives glance at page counts and mentally convert to runtime.

But is it actually true?

I first looked at this question back in 2018 when John August reached out to ask me to investigate the rule. Today’s research has a substantially larger dataset of 2,520 scripts, and I’ve been able to dig much deeper into whether the one-page-one-minute rule holds up.

A quick primer on film screenplay formatting

The formatting of a modern film script is extremely exacting, including predefined margins, layout, headings, capitalisation, and font (12-point, 10-pitch Courier). The key elements are:

• Scene heading - Setting out where the upcoming scene takes place, and sometimes a time of day/night. INT is used for interior scenes and EXT for exterior.

• Action - Descriptions of what is happening in the scene, how the characters are moving and anything else the reader/viewer needs to know which is not contained in the dialogue.

• Character - The name of the character who is about to speak.

• Dialogue - The words the character is speaking.

• Parentheticals - A few words in brackets between the Character and Dialogue, giving guidance on how the line is to be spoken. Writers are normally advised to keep these to a minimum and only to use them when vital to understanding, such as “(sarcastically)” or “(to the dog)”.

There are other elements and formatting rules, but I’m focusing on those that affect our investigation of the one-minute-per-page rule.

Is the rule correct?

For this study, I gathered screenplay PDFs for 2,520 feature films and matched them against the film’s actual theatrical runtime from IMDb.

To measure how well the one-page-one-minute rule works, I calculated a simple ratio of pages to runtime (in minutes). Perfect adherence to the rule would yield a ratio of 1, a ratio above 1 means the script had more pages than the film had minutes (i.e., the script ran ‘long’), and a value below 1 means the script was shorter than the film.

Across all scripts, the average ratio is 1.1, meaning that the typical screenplay has about 9% more pages than the film's runtime in minutes.

One page does not equal one minute - it equals about 55 seconds.

This is a much bigger deal than it might seem at first. The median script is 119 pages long and produces a 108-minute film. That’s eleven extra pages, or roughly ten minutes of ‘overshoot’ if you took the one-page-one-minute rule at face value.

Is the rule useful?

No.

There are a number of problems with using this rule in a real-world setting, including:

1. It doesn’t work for four out of five scripts

2. Screenplays are normally longer than movies

3. Different genres have very different screenplays

4. The rule falls apart for long and short scripts

5. It’s designed for Hollywood… where it’s least effective

6. Taking end crawl credits into account makes it worse!

Let’s go through each one in turn…

Problem 1 - It doesn’t work for four out of five scripts

Only 18.2% of scripts have a ratio between 0.95 and 1.05.

To put this another way, if you write a 120-page screenplay and assume it will produce a 120-minute film, there’s roughly a four-in-five chance you’ll be wrong by more than six minutes in either direction

Problem 2 - Screenplays are meant to be longer than movies

Even if the rule was true, it wouldn’t withstand the uncertainty of the journey from page to screen. A film will go through a number of stages between the first draft and the final movie.

The list below is not a strict chronology but more of an example of the many phases that can shape a film’s contents:

1. The writer(s) create multiple drafts of the script

2. Directors, producers, studios (and sometimes powerful actors) will suggest changes.

3. Changes will be made to accommodate the budget, schedule, locations, weather and force majeure.

4. The actors and director will work together to craft a performance, sometimes changing the lines.

5. The editor will edit the final scenes from what was shot.

6. “A final” cut can be affected by the requirements of censors, exhibitors, airlines and broadcasters.

Each film will have a unique set of factors affecting the final running time - very few are predictable at the script stage.

The scatter plot below shows each script as a point, with page count on the horizontal axis and runtime on the vertical. If the one-page-one-minute rule held perfectly, every point would fall on the diagonal line.

The individual variance between projects is large, as the scatter plot below illustrates. If the rule were useful, we would want the vast majority of dots to be on the dotted line.

Problem 3 - Different genres have very different screenplays

Not all genres have the same relationship between page count and runtime, to quite a large degree.

Musicals and War films are the only ones with a ratio under 1 (0.9 for musicals and 0.99 for war films). This makes sense as musical numbers eat up screen time without requiring much script space (a direction line like ‘Sarah performs the number’ can cover four minutes of screen time), while war films tend toward long, dialogue-light action sequences.

At the other end, Comedy comes in highest at 1.15. The reason that comedy scripts tend to run about 15% over on pages relative to runtime is that they rely on dialogue to a greater extent than any other genre. A rapid-fire exchange between two characters can fill a page in seconds of actual screen time, and physical comedy happens faster than they read.

The data reveals that the one-page-one-minute rule works best for action-heavy, visually driven genres and worst for dialogue-heavy ones. If you’re writing a comedy, you might want to think of it as ‘one page equals about 52 seconds’.

As a side note, I love how the genres are arranged along a spectrum from ‘visual storytelling’ to ‘verbal storytelling’. Genres that trust the camera to tell the story have lower ratios, whereas genres that rely on words have higher ones.

Problem 4 - The rule falls apart for long and short scripts

Shorter scripts tend to produce more screen time per page, while longer scripts tend to have more ‘excess’ pages.

Short scripts (60–89 pages) have a median ratio of just 0.798. These scripts consistently produce films longer than their page counts would predict. A 75-page script in this bracket typically produces a 94-minute film.

Scripts in the 90–109 page range come in at 1.0 i.e. nearly perfect, and right on the money.

The 110–130-page range (the ‘Goldilocks zone’ that most studios aim for) has a ratio of 1.093, meaning these scripts average about 9% more pages per minute.

And scripts of 151+ pages have a median ratio of 1.361, roughly a third more pages than minutes of runtime.

Longer scripts tend to be dialogue-heavy (dramas, character pieces), which we’ve already seen pushes the ratio up. They’re also more likely to undergo significant cutting in post-production, so a 160-page script was probably always going to lose material. And there’s likely a formatting effect in which writers who write long may also tend toward more verbose scene descriptions.

So if you’re writing a lean, sub-100-page script, don’t assume it will produce an equally lean film. You’ve probably got more screen time in there than the page count suggests. Conversely, if you’ve written 150 pages, the film will almost certainly come in at well under 150 minutes.

Problem 5 - It’s designed for Hollywood… where it’s least effective

When I hear about the rule, I almost never hear anyone take into account that there is no single global paper size.

In Hollywood, scripts are formatted on US Letter paper, but much of the world uses A4 paper, which is slightly narrower and slightly taller than Letter.

A4 allows for fewer characters per line but more lines per page, resulting in different page counts for identical content.

The data I have shown you thus far was only those scripts on US Letter. I also looked at a smaller subset of 351 scripts formatted on A4, where the final ratio was 1.02 - much closer to the 1.0 ratio the rule promotes.

The median A4 script is 110 pages, compared to 119 pages for Letter. That’s 7.6% fewer pages for roughly equivalent content.

If you’re a UK-based screenwriter submitting to American companies, your 100-page A4 script might produce a film of about 98 minutes - bang on target. But that same content, reformatted to US Letter paper, would come out to roughly 107 pages, making it look like it’s running seven minutes over.

Ironically, the rule of thumb works better for the paper size that didn’t invent it.

Here’s a helpful tip - if you’re writing on A4 and someone asks for a page count in ‘Hollywood pages’, multiply by roughly 1.07.

Problem 6 - Taking end crawl credits into account makes it worse!

When generating the headline data we’ve already seen, I swithered over how to account for end crawl credits. In the end, I figured people most often use the rule to describe a whole screenplay rather than any one individual page. Therefore, the data we’ve seen uses raw IMDb running times, which include credits.

I am working on a larger study of credits that will be released in the coming months, but right now, I could sample 837 movies to get a rough length for crawl.

The median end credits length across all the movies I studied is 4.3 minutes.

They’re getting long - much longer! In the 1970s, credits averaged 2.0 minutes. By the 1990s, that had risen to 3.7 minutes. In the 2020s, it’s 5.2 minutes. Credits have roughly tripled in length over fifty years, and every extra minute of credits is a minute of runtime that has no corresponding page in the screenplay.

What happens to the one-page-one-minute rule when you strip credits out and measure pages against story time only? The rule is even less accurate than the main analysis suggested. One page equals not 55 seconds of film, but closer to 54 seconds of actual story.

Does the rule help or hurt screenwriters?

Both.

The rule is obviously a quick and easy way to see if you’re roughly in the right ballpark when it comes to the total running time of your screenplay.

In addition, if you’re hoping to hit certain story beats by a particular moment in the film, then knowing you’re reaching it by page [x] can help you focus your writing.

Finally, as a writer primarily concerned with crafting good characters, stories, and emotions, the exact running time is largely irrelevant. I’ve seen 90-minute movies that have dragged and three-hour ones that flew by.

The problem comes when the rule is used against screenwriters, not by them.

I have heard no end of stories of writers being told to cut their scripts based purely on page count. Hollywood screenwriter John August puts it like this:

Unfortunately, too many folks in the film and television industry have internalized one-page-per-minute as an axiomatic Truth. So any script that is longer than 120 pages is automatically perceived as being too long.

Indeed, some studios’ contracts specify that the writer may not deliver a script longer than 120 pages. Screenwriters waste time making tiny edits with the goal of moving page breaks to bring their scripts under this artificial limit. It’s pointless busy work.

Ultimately, running time is a factor of film editing. Scenes get dropped in post, and it’s very hard to anticipate these changes when looking at a script in preproduction.

Interestingly, I think we can see the effect of the rule in action.

Scripts from the 1930s averaged about 45% more pages than the film had minutes. Early screenplay formatting hadn’t been standardised; scripts tended to include much more scene direction and camera instructions, and the relationship between page and screen hadn’t been fully understood.

The ratio has been on a remarkably steady downward trajectory ever since. The 1960s at 1.14. By the 1980s, it had fallen to 1.13.

The 1990s saw a significant step down to 1.08, and the 2000s landed at 1.110, while the 2010s came in at 1.064.

The 2020s represent the first decade in our dataset in which the median ratio has fallen below 1.0, at 0.99.

I put this down to a few factors:

1. Working to rule. As the rule became law, deviations from it were incrementally punished. In response, writers would aim to meet these requirements as closely as possible, saving the odd word here and there.

2. Screenwriting software was a huge leap towards better formatting standardisation (Final Draft was first released in 1990), and increasingly rigid studio expectations about script length.

3. “Show, don’t tell”. Modern screenwriting advice encourages leaner, more visual writing. ‘White space on the page’ has become a mantra in screenplay workshops, and the verbose, novelistic screen directions of earlier decades are frowned upon.

4. Something something streaming. Many films produced for streaming platforms are shorter than traditional theatrical releases, which intuitively suggests they would have some effect.

Conclusion

The one-page-one-minute rule is not a bad starting point. It’s certainly more useful than having no rule at all. And it’s become more accurate over time, as formatting has standardised and screenwriting conventions have converged.

But it’s also very imprecise. Fewer than one in five scripts actually hits the target range. Genre, era, page count, and even paper size all shift the ratio in predictable ways. The ‘rule’ is less a rule and more a midpoint around which the industry ends up, with substantial variation in every direction.

If I had to offer a revised rule for 2026, it would be this:

One page of a properly formatted screenplay equals approximately one minute of screen time, give or take 20%, depending on genre, era, and paper size.

Sure, it’s not as catchy as the original, but it’s a good deal more accurate.

And if you’re writing a comedy on A4 paper, you can safely ignore everything I’ve just said, because your experience will be entirely different from someone writing a war film on US Letter.

Notes

Today's research looked at 2,520 movie screenplays in US Letter format and 351 in A4 format. There are plenty of guides online on how to format a screenplay. It’s also worth noting that film scripts are formatted differently from TV, radio and theatre plays.

I tried to find the “final” script version of each movie, but this is inherently imprecise as I couldn’t check every line against the final film.

Page counts are taken from the PDF. Some scripts include title pages, cast lists, or appendices that inflate the number. I cleaned these where obvious, but some noise will remain.

Runtimes come from IMDb, which typically reflects theatrical runtime, including credits. Some scripts may correspond to different cuts (director’s cut vs. theatrical release). This is a known source of noise, but is unlikely to affect the overall patterns.

Genre assignments come from IMDb, meaning that each film can belong to multiple genres (a film might be both ‘Comedy’ and ‘Romance’). Some genres have small sample sizes. Musical (19 scripts) and Western (30 scripts) should be interpreted cautiously, though their ratios make intuitive sense. The bigger genres, such as Drama (1,422), Comedy (849), and Action (608), provide much more statistical confidence.

Decade sample sizes. The 1930s (23 scripts), and 1940s (33 scripts) have relatively small samples, so those figures should be treated with caution. The patterns become much more reliable from the 1970s onwards.

Correlation vs. causation. The genre and decade patterns show correlation, not causation. A comedy doesn’t have a high ratio because it’s a comedy - it has a high ratio because comedies tend to be dialogue-heavy, and dialogue is dense on the page. The underlying driver is the type of content, not the genre label.

I should note that this isn’t a perfectly random sample of all films ever made. It skews toward well-known titles with publicly available scripts, which means big studio releases are over-represented, and low-budget independent films are under-represented. That said, it’s a large enough dataset that the broad patterns should be reliable, even if the exact numbers might shift with a different sample.

Also, in past research projects, I have reviewed over 12,000 screenplays from aspiring writers, and I don't see any differences in formatting between the two datasets.

Epilogue

I’m always learning and getting access to better tools and data. I first tackled this topic in 2018, analysing 761 scripts. I was pleased to see that the results were very similar. My 2018 study found 22% of scripts within the ‘perfect’ range, and this 2026 update finds 18.2%. The median page count has held steady at 119.

The main addition in this study is depth, with genre breakdowns, decade analysis, page-count brackets, A4 comparison, and a larger sample size, all of which add nuance that the 2018 study couldn’t provide.

Despite more than tripling the sample size, the core numbers have barely shifted. This gives me confidence that the one-page-one-minute rule, while imprecise, is at least a stable imprecision. It’s not fluctuating wildly from sample to sample, i.e. the typical script really does run a little longer than its film, and the ‘overshoot’ is consistently in the range of 9–10%.

I’ll be uploading new video essays every couple of weeks throughout 2026, so make sure you’re subscribed on YouTube to get notified when another drops (also, subscribing will really help me out with the Algorithm Gods, so it would mean a lot to me - thank you!).

https://www.youtube.com/@TheFilmDataScientist

If you want to know more about the channel, here’s a quick primer:

• Movies are becoming more simplistic with less sex.

• Lighting is flatter and human faces are larger on screen.

• The pacing has become more intense, with less quiet time and more dialogue

• We’re teaching different moral lessons and romantic love is giving way to other forms of connection

Today we’re going sonic. Not, not that one - we’re moving on to the audio of movies.

I built a system to identify every sound event from almost 2,000 action movies made over the past 75 years. We’re tracking gunfire, vehicles, music, speech, explosions, animal sounds, impacts, and more.

I was a little surprised by what I found. Let’s start with perhaps the easiest to guess…

Action films are getting audibly busier

The total number of detected sound events per minute in action movies has been rising steadily.

A modern action film packs in roughly 15 distinct sound events per minute. In the 1970s, it was closer to 13. That might not sound like a dramatic shift, but it adds up. Over a two-hour film, that’s about 240 extra sound events competing for your attention.

This is partly a technology story. Multi-track digital recording, surround sound, and modern DAWs make it easier to layer sounds.

But it also reflects a shift in audience expectations. Modern action films are denser, faster, and louder because audiences have been trained to expect that. A mid-1970s-era soundscape in a 2026 action film would feel eerily sparse. (I wrote about having exactly this feeling when I watched The Conversation (1974) last year in Has the way movies use dialogue changed?).

Gunfire is up 44%

A major finding is that gunfire events in action films have increased by 43% since 2000. Pre-2000, a typical action film averaged 0.196 gunfire events per minute. Post-2000, it’s 0.280.

I broke gunfire down into subcategories: single gunshots, automatic weapons fire (machine guns and fusillades), explosions, and artillery.

This shows that explosion events have nearly doubled since the 2000s, rising from 0.059 per minute to 0.111 per minute in the 2020s. Automatic weapons fire has been rising steadily since the 1970s, as have single gunshots (i.e. the classic revolver or rifle shot), albeit more modestly.

What’s happening here is an escalation of intensity. The modern action film has both more gunfire and more sustained gunfire. The deliberate, spaced-out gunplay of a 1970s thriller has given way to the continuous barrage of a John Wick or Extraction.

My first thoughts as to why this might be happening are:

1. Video games. Many modern action sequences are structurally similar to first-person shooters, with continuous fire and constant explosions.

2. Economics. Digital muzzle flashes and CGI explosions are vastly cheaper than practical effects, making it easy to extend and intensify action sequences.

3. Franchises. Sequels demand escalation, with each needing to top the last.

The list of the most gunfire-heavy action films is dominated by war films. This makes sense as war films are essentially extended combat sequences. But the presence of films like Olympus Has Fallen, Extraction II, and Rambo (2008) shows that non-war action films are approaching war-film levels of gunfire density.

Only one of the ten ‘most gunfire-heavy action films’ was made before 2000 - Hamburger Hill whcih was the most gunfire-intensive film I tracked.

What’s very interesting is that last year, I discovered that fewer lead characters are dying of gunshots. The only logical conclusion is that the aim of action movie villains is getting considerably worse.

Music is the biggest single change

If I had to pick the single biggest shift in action film sound, it’s the music.

Music scoring in action films has risen by 9% since 2000, from 5.04 to 5.48 events per minute. That’s the largest absolute change of any sound category (+0.44 EPM).

My take here is that the franchise era demands wall-to-wall scoring. Marvel films, DC films, and their imitators use continuous orchestral underscoring to maintain emotional intensity. The 1970s approach of long stretches of diegetic sound punctuated by targeted musical cues has been replaced by a model where the score rarely stops.

Modern action films are, in a very literal sense, more like musicals than their predecessors. It is perhaps no surprise that some of the most exciting action films of the past few years have come from Bollywood cinema (RRR, Pathaan, Jawan, War, Baaghi 2, etc), which is taking more cues from musicals than Hollywood action films.

There’s also a loudness dimension here. Music events that register on the classifier tend to be prominent in the mix. The trend toward louder, more compressed masters (i.e. the ‘loudness war’ that started in music and migrated to film) means music is more likely to be detected even in scenes that are nominally dialogue-driven.

The car chase is dying

Here’s one I didn’t expect. Vehicle sounds in action films have declined by 16% since 2000. This neatly tracks with my past research showing that the fastest declining cliché in movies is “Follow that [vehicle]”.

In the 1970s, action films averaged 1.68 vehicle events per minute. In the 2020s, it’s 1.13.

Digging into the data, we can see that fewer cars are a major cause of this decline.

Cars still remain as the dominant vehicle sound, but two specific subcategories tell a fascinating story.

• Helicopters peaked in the 1980s at 0.02 EPM and have been declining since. This is almost certainly the Cold War and Vietnam War’s influence on 1980s cinema, with Apocalypse Now, Blue Thunder, Rambo, and their imitators filling the soundscape with rotors. Modern action has replaced the helicopter with drones (which are sonically quiet) and CGI spacecraft.

• Emergency vehicles have declined as the cop thriller migrated from cinema to television. The kinds of films that once filled multiplexes with sirens and screeching tyres (e.g. Dirty Harry, The French Connection, Lethal Weapon) are now eight-episode prestige TV series.

The broader story is a shift in the grammar of action cinema. The car chase was the centrepiece of 1970s and 1980s action, whereas modern action films centre on hand-to-hand combat, gunfights, and superhero set pieces.

Fast & Furious is the exception that proves the rule, and even that franchise has moved increasingly toward absurdist spectacle rather than traditional car chases.

Four sounds that tell a story

Zooming in on specific sound events reveals stories that the broader categories mask:

1. Helicopters are a Cold War sound. They peaked in the 1980s and have steadily declined. Today’s filmmakers have replaced them with drones (quiet), CGI spacecraft, and ground-level action. The helicopter was the signature vehicle of the Vietnam War film and the 1980s military thriller, but that era is over.

2. Explosions have nearly doubled since the 2000s. The Marvel Cinematic Universe and the ongoing escalation of franchise third acts are driving this. Modern climaxes are essentially continuous sequences of explosions.

   

3. Automatic weapons fire has been rising steadily, reflecting a shift from revolvers and rifles to fully automatic weapons. John Wick fires more rounds in a single scene than Dirty Harry fired in an entire film.

4. Speech is remarkably stable. Action heroes talk about as much now as they did in the 1970s, roughly 6.0 speech events per minute. Despite the increase in gunfire and explosions, dialogue density hasn’t been squeezed out. Modern action heroes simply talk between the explosions rather than instead of them.

Is this just an action film thing?

I ran my detection system on all types of movies, allowing us to see how many of the changes we’ve seen are action-specific.

Interestingly, the average action movie now has more dialogue events than non-action films. Both cohorts have seen more spectacle in the past few decades, but action movies have been the more talkative of the two over the past ten to fifteen years.

The rise in gunfire is also seen in other genres. Fantasy films have seen a marked increase this century, as the genre shifts from Middle-earth to the Marvel Cinematic Universe.

Only romantic films have seen a massive decline, and as the romance genre has been disappearing from our screens (from 34.8% of releases in 2000 to just 8.6% in 2024), our cinemas are increasingly filled with gunfire.

Despite the rise in fantasy and horror gunfire, action still tops the list of gun-heavy genres.

Action set pieces have infiltrated comedies (21 Jump Street), dramas (Sicario), and even animation (Spider-Verse). The boundaries between genres are blurring, and the sound is blurring with them.

So what’s changed?

If you could listen to the average action film from 1985 and the average action film from 2026 side by side, the differences would be clear but probably not what you’d expect.

• The modern film would be sonically denser, with more events competing for your attention every minute.

• It would have more gunfire, and specifically, more sustained automatic fire and explosions.

• It would have significantly more music, nearly wall-to-wall.

• It would have noticeably fewer vehicle sounds, car chases, helicopters, and sirens.

• And yet the amount of dialogue would be almost identical.

The human voice, it turns out, is the constant. Everything else (i.e. the gunfire, the vehicles, the music, the explosions) is the variable.

Sound designers and directors have been reshaping the audio landscape of action cinema for decades, but they’ve done it around a fixed point of around six events of the human voice per minute.

Notes

Today’s research is based on the audio from 1,797 action feature films released between 1950 and 2025, identified by their IMDb genre tags.

I used a PANNs CNN14 neural network trained on Google’s AudioSet taxonomy (527 sound classes). The classifier processes audio in 10-second windows with 2-second overlap, detecting and categorising sound events which are then aggregated into 10 broad categories.

I calculated events per minute (EPM) as the count of detected events in each category divided by the film’s total audio duration in minutes. This gives an absolute measure of sound density independent of other categories.

Vehicle subcategories were grouped from AudioSet classes:

• Cars (car, racing, skidding, tyre sounds)

• Emergency vehicles (police/fire/ambulance sirens)

• Aircraft (helicopters, planes, jets)

• Boats

Gunfire subcategories:

• Single gunshots

• Automatic fire (machine gun, fusillade)

• Explosions (explosion, boom, eruption)

• Artillery

The classifier is not perfect. It can misidentify sounds, particularly in dense mixes where multiple sources overlap. Results should be treated as indicative of broad trends rather than precise measurements. All statistical comparisons use Spearman rank correlation; changes described as significant meet p<0.01.

Films must have a minimum audio duration of 10 minutes to be included. Genre labels are from IMDb and are non-exclusive (a film can be Action, Adventure, and Sci-Fi simultaneously).

His success reminded me of something I was looking into a few years ago, but dropped the thread on. Namely, how many people have taken on all four major roles in movies before.

I was surprised by the lack of research on the topic so today, I would like to propose a word to describe the phenomenon and take you through the numbers

Meet ‘Egocore’

Here’s my new definition:

Egocore. A feature film where one credited individual is simultaneously a credited writer, a principal producer, the director, and a top-billed on-screen performer (i.e. not a cameo appearance).

My classification was initially inspired by watching movies by the indescribable, incomparable filmmaker Neil Breen. I have seen all of his movies, and I’m still not sure what they’re about or what’s going on. But as outsider art goes, it’s amazing.

To say that Neil takes multiple roles would be an understatement. Across his last seven movies, Neil has received 56.4% of all crew credits.

A few years ago, I attended the packed UK premiere of Cade: The Tortured Crossing at the Prince Charles cinema in London. The audience was so committed to Neil's unique style of cinema that, when the end credits rolled, there were loud cheers for each of Neil’s credits and boos for anyone else.

Neil’s work got me thinking about other filmmakers who were wearing multiple hats on their productions. Few take on as many crew roles as Neil does, so I thought it was best to limit it to people who “just” write, produce, direct, and star in a movie.

I studied 38 million credits across 614,342 movies made since 1940 to find two types of productions:

1. Strict egocore, where the person took on all three creative roles and was in the top 3 of the cast.

2. Broad egocore, where they got slightly weaker credits in some respects (e.g., not a full writing credit but ‘story by’, or only an Executive or Associate Producer).

I found 11,827 strict egocore films and 16,492 broad egocore films.

So roughly one in every 37 films qualifies as egocore under the broad definition, or about one in 52 under the strict one. That’s more common than I thought.

The rise of Egocore

Egocore is not a wholly new phenomenon, but it has become dramatically more common.

In the 1940s, about 0.14% of films qualified. By the 2010s, that had risen to 2.19% under the broad definition (i.e. a more than thirteen-fold increase).

This acceleration aligns neatly with the democratisation of filmmaking tools. Digital cameras, affordable editing software, and online distribution all made it feasible for a single person to fill every role.

Here’s my breakdown of the major Egocore eras:

• 1940s–1960s - The Chaplin era. Egocore was vanishingly rare, with Charlie Chaplin as its main practitioner. His four egocore films from this period (The Great Dictator, Monsieur Verdoux, Limelight, A King in New York) are among the best-known movies in my dataset. At this point, egocore required genuine industrial clout. You needed to own your own studio or be so famous that nobody could say no - Chaplin had both.

• 1970s–1980s - The auteur window. New Hollywood loosened the studio system enough for a handful of filmmakers to consolidate control. Mel Brooks wrote, directed, produced, and starred in Blazing Saddles and History of the World: Part I. Dev Anand was building his egocore career in Indian cinema, but overall egocore was still rare, and still the province of established names.

• 1990s - The indie explosion. The Sundance generation and the rise of affordable 16mm and early digital cameras made independent filmmaking viable for more people. Egocore doubles in size and starts to diversify beyond the A-list.

• 2000s–2010s - The digital revolution. This is where the curve bends sharply upward. DSLRs, Final Cut Pro, and platforms like YouTube and later streaming services created an ecosystem in which a single person could genuinely handle every aspect of production. The egocore rate climbs from under 1% to nearly 2% of all films. Tyler Perry releases the bulk of his Madea egocore films and Neil Breen emerges as an unlikely cult figure.

• Current decade - Unknown future. The rate appears to plateau or dip slightly, though the decade is only half done. The pandemic temporarily disrupted production, and the post-pandemic landscape has been dominated by streaming commissions, in which collaborative production structures are the norm. Whether egocore continues to c

• limb or has reached its natural ceiling remains to be seen.

The most-watched egocore films

Statistically speaking, you’ve probably seen a few egocore movies and never even knew it. Zoolander is a great example, where Ben Stiller is firing on all cylinders in each of the four required creative roles to qualify.

Egocore by genre

Not all genres are equally prone to egocore. I looked at the egocore rate by primary genre for genres with at least 1,000 films in the database.

While comedy took all seven of the top spots on the ‘most popular’ chart above, sci-fi and horror have the greatest relative production overall.

Egocore is often the choice of the cash-strapped filmmaker

Egocore is overwhelmingly a low-budget phenomenon.

The egocore rate holds fairly steady for films budgeted under $50 million, but then drops off a cliff, with only a couple of egocore movies costing more than $50 million.

The more money involved, the more stakeholders are involved, and the harder it is for any single person to maintain control of all four key roles.

Who are these people?!

Men. Roughly four to one (82.3% vs 17.7%).

I know what you cynics are thinking - that’s typical of men’s overconfidence in being able to take on any task (46% of men believe they could manually land a passenger airplane if they really had to in an emergency).

But to that I say, you’re not thinking cynically enough. Film has historically been a hugely sexist industry, and so when you combine roles, maybe you also combine barriers.

Over the last century, acting has been the closest to parity at 39.5% women. Among the top creative roles, producing has been 27.1% women, writing 20.2%, and directing (the role most associated with singular creative authority) sat at just 17.0% women.

Full egocore comes in at 15.4% women. That’s almost the same as the directing rate, which suggests that it’s the directing barrier, rather than anything specific to egocore, that’s the bottleneck for women.

The trend is moving towards gender parity, albeit slowly. In the 1940s, roughly 5% of egocore filmmakers were women. By the 2020s, that figure had risen to about 20.2%.

So it’s not that egocore is uniquely male-dominated, but rather that it inherits the gender imbalance of its most male-skewed component - directing.

Egocore’s heroes

I’m not sure what we should call the filmmakers behind such films…

• Egoteur?

• Egocorists?

• Singularists?

• Uniplexers?

• Polymakers?

• Egocorepreneurialists?

Despite the prevalence of the films, of the people who have made at least one egocore film (using the broad definition), the vast majority, roughly 89%, have done it exactly once.

Only about 90 people have made four or more egocore films, and just 19 have made seven or more. Egocore, for most filmmakers, is a one-off experiment.

The serial practitioners are a tiny, prolific minority. These heroes include:

• Charlie Chaplin is the original egocore OG, proving that egocore, in the right hands, can produce truly great art.

• Tyler Perry is prolific, commercially savvy, and unapologetically populist. His egocore output is remarkable both for its volume and its consistency, meaning that his audiences know exactly what they’re getting, and they keep coming back. (More on Tyler here What filmmakers can learn from Tyler Perry’s incredible self-made empire).

• Neil Breen has become a cult phenomenon precisely because of his egocore commitment. On his early classic Twisted Pair, Neil is credited as the film’s director, writer, producer, cinematographer, director of photography, editor, hair stylist, makeup artist, special makeup effects artist, production manager, set designer, sound editor, special effects artist, lighting designer, casting, wardrobe, locations manager, music director, craft service, props, aerial rigging, legal accounting administration, and stunt coordinator. Only two other people received a crew credit on the movie.

• Xavier Dolan is perhaps the most critically acclaimed modern egocore practitioner. He burst onto the scene at Cannes aged 19 and went on to write, direct, produce, and star in several films that were taken seriously by the festival circuit.

• Dev Anand is a reminder that egocore isn’t just a Western phenomenon. With 12 egocore films spanning decades of Indian cinema, Anand was one of the most prolific egocore filmmakers in any national cinema.

• Nanni Moretti represents the European auteur tradition of egocore in which his films are cerebral, personal, and deeply rooted in national identity.

• Mel Brooks demonstrates that comedy may be the genre best suited to egocore. When a comedian writes their own material, directs their own timing, and performs it themselves, the result has an internal coherence that’s hard to replicate by committee. Brooks’s egocore comedies are among the most beloved comedies ever made.

Notable near-egocore

Given the tough requirements of fulfilling all major creative roles simultaneously, it’s perhaps not surprising that many of the people we regard as singular auteurs still fall short of the four-out-of-four required to join this elite group.

Well-known filmmakers who are egocore runners-up include:

• Woody Allen is arguably the most famous near-egocore filmmaker. He wrote, directed, and starred in numerous films without a formal producer credit on any of them. In practice, Allen exercised a level of creative control that was functionally egocore, but the credits don’t reflect it.

• Clint Eastwood shows a similar pattern, with numerous films he directed, produced, and starred in, but never took a writing credit.

• Sylvester Stallone wrote, directed, and starred in several Rocky and Rambo films but typically lacked a producer credit.

I ran the numbers on job overlap for the four roles.

• By far the most common three-role combination is Writer-Director-Producer (missing a top-billed acting credit), accounting for 81% of all near-egocore instances. These are the behind-the-camera auteurs who control everything except providing a performance themselves.

• The second most common is Writer-Director-Actor (missing Producer), with 8% of cases. These filmmakers write, direct, and star but don’t take a producing credit, as Woody Allen does.

• Writer-Producer-Actor (missing Director) accounts for 6% of near-egocore misses. This is the actor-creator who develops their own projects but hands the directing to someone else. A sensible approach, frankly, as directing yourself while acting is notoriously difficult.

• Rarest of all is the Director-Producer-Actor (missing Writer), occurring in just 4% of cases. These are filmmakers who control the production and star in it but work from someone else’s script, such as Clint Eastwood.

The hierarchy makes intuitive sense to me. The ‘easiest’ role to hold onto behind the scenes is producing (maybe the hardest job, but certainly the most liberally credited). While the hardest role to combine with the others is acting, because it requires a fundamentally different skill set and a willingness to be in front of the camera.

Are egocore films any good?

I’m sure by now you’re wondering whether egocore films are worth seeking out or avoiding. Is all that control a recipe for self-indulgent disaster?

The short answer is that it’s basically a wash. Across most decades, the average IMDb rating for egocore films is within a fraction of a point of non-egocore films. In some decades, egocore edges ahead; in others, it falls slightly behind. There is no dramatic ‘ego penalty’ in the aggregate.

But the genre-level picture is more interesting.

Egocore comedies rate notably higher than non-egocore comedies. The same is true of comedy-dramas. It seems that personal comedic vision benefits from singular control. When the person writing the jokes is also delivering them, there’s an internal coherence that committee-written comedy often lacks.

What have we learned?

Egocore is real, it’s growing, and it’s more nuanced than the caricature of a vanity project suggests.

The data tells us several things worth noting:

1. Egocore is mostly a micro-budget phenomenon. The vast majority of egocore films are made for under $5 million. The person wearing all four hats is usually not doing so because they have a studio behind them. They’re doing it because there’s nobody else to do it. This is ego-as-necessity as much as ego-as-vanity.

2. The quality distribution is surprisingly normal. Egocore films aren’t dramatically better or worse than their non-egocore equivalents. The idea that handing one person total control inevitably produces either genius or garbage isn’t supported by the data. Most egocore films are simply... fine.

3. Genre matters more than ego. Whether an egocore film works well seems to depend more on what genre it’s in than on the fact of its egocore-ness. Personal comedic visions benefit from singular control; complex action sequences and horror set-pieces may suffer.

4. The near-egocore club is as interesting as the full one. Woody Allen, Clint Eastwood, and Sylvester Stallone are all near-egocore by the formal credits but clearly exercise egocore-level control. The credits system doesn’t always capture the reality of who’s actually running the show.

5. The pathway to egocore almost always starts behind the camera. The most common three-role combination is writer-director-producer (missing actor), with 99,829 instances. The last step into full egocore (i.e. stepping in front of the camera) is the one most filmmakers never take.

Epilogue

I reached out to a few of the filmmakers I’ve mentioned today to share their perspective. Neil Breen put it like this:

I have done 7 indie ‘B’ theatrical feature films. I was the producer, director, writer and star of each film.

I own the copyrights to the films and scripts. I own 100% of the legal rights to the films. I self-funded my films. Each film was profitable which enabled me to do the next film.

I’m a team player and look forward to collaborating but I live in an area which does not have a pro film community and therefore the film sources and resources are very limited.

My passion is film making. So rather than talk about it or wait for someone to help me, I do it myself. I’m very experienced, creative and skilled in doing the work myself.

Notes

Today’s analysis relies on tracking people’s credits, not the work they did. Some filmmakers exercise egocore-level control without formal credits in all four categories (Woody Allen being the most obvious example). Others may have credits that overstate their actual involvement.

Billing order as a proxy for ‘starring’. I used top-3 billing as the threshold for being a ‘star’ actor. This is a reasonable proxy but not perfect, as some films list actors alphabetically, and billing politics can distort the picture.

The ‘strict’ writer definition counts only screenplay and ‘written by’ credits. The ‘broad’ definition adds story credits. Some filmmakers contribute to the script without a formal writing credit (and vice versa).

Budget information is a mix of confirmed budgets and estimates using a model I’ve been working on for a few years. I have only done this for films made this century, so charts and data are all for 1940-2025 unless they reference budgets, which means they are 2000-25.

Gender estimation is imperfect and could miscategorise some non-Western names and non-binary individuals. The figures here should be treated as directional estimates, not precise counts. But the overall pattern (i.e. heavy male dominance of egocore) is robust to any reasonable error margin.

I haven’t engaged much with writing academic papers over the years due to time constraints. But it strikes me that this is a rich seam for someone to mine. So if you’re an academic who wants to do a paper on this, reach out.

In a world of multiple screens, endless distractions and an attention span of… wait, what was I saying?

Oh, yeah, intensity is increasing because people can’t focus on… you know, the thing.

In a recent interview on The Joe Rogan Experience, Matt Damon articulated this point of view far more eloquently than I have. He explained that when they made their recent film The Rip for Netflix, the streamer gave them very specific notes on how to structure their story:

The standard way to make an action movie that we learned was you usually have three set pieces… and now [Netflix is] like, ‘Can we get a big one in the first five minutes? We want people to stay tuned in. And it wouldn’t be terrible if you reiterated the plot three or four times in the dialogue because people are on their phones while they’re watching.’”

So Netflix certainly feel that movies are (or should be) paced differently than they used to be. Are they right? Are movies more intense? Is the way movies tell stories changing?

There are a few places we might look to get to the heart of this, including:

1. Stories - I.e. What stories are being told and the types of characters whose journeys we follow. See: What are the most common moral lessons within movies?

2. Scriptwriting - How the stories are being told and the choices the writers are making in the dialogue. I touched upon writing changes here Are movies becoming more simplistic?

3. Sounds - I might be getting old, but it certainly feels like movies are getting louder.

4. Cinematography - Last year, I looked into whether modern movies use flatter lighting than older movies. I won’t spoil the result, as you can read the whole thing here: Are modern movies using flatter lighting than in the past?

5. Sins - In past research, I have shown that the levels of drug use, violence, and swearing have not changed in recent decades, but the levels of sex have fallen by around 40% since 2000. More at Why is sex in movies declining?

Today I’m going to tackle the second on that list. Namely, are stories being told in a more intense way than they used to be?

To find out, I analysed the dialogue of 70,985 English-language feature films released between 1940 and 2022. In each movie, I mapped a number of linguistic high-arousal signals (i.e., fear, anger, urgency, profanity, punctuation pressure, and emphasis), scoring them and noting their timing within the movie.

This gave me a deep insight into how movies build, sustain and release dramatic intensity.

Here’s what I found…

Have movies actually become more intense?

The story of the 20th century was indeed that “movies are getting more intense”, but in the past few decades, things have actually declined ever so slightly.

The most intense year for movies was 1991.

Below is the per-year average intensity of the movies I studied. We see a strong rise from the mid-1940s through the mid-1960s. Intensity drops slightly towards the 1970s, then resumes a steady rise, peaking in the early 1990s. Since then, the picture is mostly a gentle decline, with a brief rise in the early 2010s.

The chart above looked at the average accross the whole movie, whereas perhaps we would like to study the most intense point in each movie. When we do that, we see a much flatter journey. Four decades (the 1960s, 1980s, 1990s and 2000s) shared the same level of intensity extremes.

Since around 2010, the ‘most intense’ part of the average movie has been less intense than in the past half-century.

So we’re not seeing the kind of results one would expect if it were true that “recent movies have been getting more intense”.

So something else must be driving the perception.

Did peaks get bigger, or did films just stop calming down?

The more interesting change is not how high films climb but how long they stay there.

Between 1940 and 2022, the share of runtime spent in high-intensity dialogue rose by 21%.

We, as the audience, are spending more time in a high state of intensity (well, technically the characters are, but I’m going to assume we’re matching their feelings).

Another way to graph the same effect is to look at how movies ease off the intensity in the middle. This “downshift index” fell by roughly 35% over the past eighty years.

This means that older films often follow a familiar rhythm - namely an early destabilisation, then a spike, followed by a middle section that cools. But many contemporary films behave differently. Once the pressure is introduced, it sticks around.

So movies have become more intense, not because their peaks have risen but because the spaces between them have reduced.

When did this shift happen?

Regular readers will see a familiar pattern in the 20th-century data. In a number of past articles, I have noted the idea that “Something weird happened with movies in the 1970s”. We’ve seen this in:

• How often do lead characters die in movies? In the 1970s, lead characters became significantly more likely to die on screen, reflecting a shift toward darker, more morally ambiguous storytelling.

• Has the way movies use dialogue changed? 1970s dialogue became looser, more naturalistic and less formally structured, moving away from mid-century theatrical polish toward a grittier, more conversational style.

• Can we measure if movies have become “too woke”? The 1970s were a period of more politically and socially controversial movies, marking a sharp rise in overtly divisive themes.

And we see it in today’s research, too.

The single largest year-on-year jump in the overall dramatic intensity index occurs in 1971.

While 1971 is dramatic, it’s probably better to describe the 1970s as a traditional era rather than to attribute all the change to a single year. Pre-1970 films show stronger mid-film downshifts and lower high-intensity occupancy. Since 1971, the baseline has lifted and remains elevated.

What this means for how films feel

To sum up, when audiences say that films feel more intense today, they may not be reacting to bigger explosions or louder climaxes. Instead, they may be reflecting the change in rhythm.

Mid-century films were often comfortable alternating between pressure and relief. Emotional spikes were followed by cooling phases, and while the stakes remained, the tone widened. Not everything was always on the line at every moment in a way that left the characters always maximally charged.

Contemporary films don’t want to let go of you, even for a moment. Once pressure enters the story, it tends to persist, via more intense middle acts and constant reminders of the emotional stakes.

Cinema today does do not want you to… just… breathe.

Notes

Thank you to everyone who got in touch after the Matt Damon interview to ask me to look into the data behind what he was saying. I am particularly grateful to Jayson Abalos, who gave a detailed accounting of his theory, including (correctly) predicting that the 1970s were a key inflexion point for the trend.

Today’s analysis is based on 70,985 English-language feature films released between 1940 and 2022. I have reported the result at the year level using medians.

The measure I used is a subtitle-surface proxy for dramatic intensity. It captures dialogue-level emotional pressure, rather than cinematography, editing, music, or performance. Intensity is constructed from linguistic and formatting signals within subtitles, including high-arousal vocabulary (fear, anger, threat, urgency, profanity), punctuation pressure (such as exclamation and question marks), and all-caps emphasis.

To compare films of different lengths structurally, I normalised them into 20 equal-run-time bins, representing narrative progress from beginning to end. I then worked out that an intensity score between 0 and 1 is computed for each bin, yielding a 20-point intensity curve for each film.

From these curves, I had two broad types of metrics:

• Amplitude measures included average intensity across runtime, upper-percentile intensity (a proxy for peak pressure), and the share of runtime spent above a high-intensity threshold.

• Structural measures captured rhythm and persistence, including the “downshift index” (how much films cool in the middle after early destabilisation), the “ratchet index” (whether pressure builds again toward the end), the number of high-to-low resets, autocorrelation between adjacent bins, and the longest uninterrupted high-intensity stretch.

I calculated a composite dramatic intensity index as a standardised z-score across those components.

Epilogue

This is the second time I have been pushed to conduct data research based on something Matt Damon said. Previously, I fact-checked his thoughts on changing movie budgets in Has the mid-budget drama disappeared?

I guess I’m lucky that, in both cases, it’s only taken me a day or so, and not the combined $900 billion that his characters have apprently cost humanity. According to Kynan Eng, this breaks down as:

• Courage Under Fire (Gulf War 1 helicopter rescue): $300k

• Saving Private Ryan (WW2 Europe search party): $100k

• Titan A.E. (Earth evacuation spaceship): $200B

• Syriana (Middle East private security return flight): $50k

• Green Zone (US Army transport from Middle East): $50k

• Elysium (Space station security deployment and damages): $100m

• Interstellar (Interstellar spaceship): $500B

• The Martian (Mars mission): $200B

• TOTAL: $900B plus change

It’s just like the original research but dumber (read: more entertaining).

You can read more on the topic here.

I’ll be uploading new video essays every couple of weeks throughout 2026, so make sure you’re subscribed on YouTube to get notified when another drops (also, subscribing will really help me out with the Algorithm Gods, so it would mean a lot to me - thank you!).

https://www.youtube.com/@TheFilmDataScientist

Here are some past videos from the channel:

And to help explain the channel, I also created a short channel trailer:

I have written a few times about how patents can be an excellent way to see which technologies will soon affect / improve / disrupt the industry.

This has led a number of readers to reach out and ask more about the technology in Ben Affleck’s AI company that would make Netflix want to buy it.

Today I’m going to go through the background, the patent and what it might mean for the film business in the near future.

If you want more context on today’s topics, here are a few recent articles which may help:

• What Netflix’s patents reveal about the future of watching movies

• What Amazon’s patent filings reveal about how filmmaking may change in the future

• How AI tools are already changing the jobs of film professionals today

• Netflix owns some pretty strange trademarks, and they reveal a lot about the company

The background - Netflix bought Ben Affleck’s AI filmmaking company

Netflix announced a few days ago that it has acquired InterPositive, an artificial-intelligence filmmaking company founded by Ben Affleck.

InterPositive was founded in 2022 and operated largely in “stealth mode” under the corporate entity Fin Bone LLC. The company says it’s developing tools intended to support filmmaking rather than replace it, focusing on production workflows such as relighting scenes, removing wires from stunt shots, correcting continuity issues and enhancing backgrounds in post-production.

The company itself is fairly small, with just 16 engineers, researchers, and creative staff, but in the world of AI start-ups, that doesn’t mean it couldn’t be cooking up industry-changing tech.

Netflix hasn’t revealed how much it paid, but we do know that all the assets, etc., and staff will now be a part of Netflix, and Affleck will remain as senior advisor.

The one specific thing we do know about is a patent.

The patent - US 12,511,904 B1

In December 2025, the United States Patent Office granted US Patent 12,511,904 B1, titled:

“Method, System, and Computer-Readable Medium for Training a Captioner Model to Generate Captions for Video Content by Analyzing and Recognizing Cinematic Elements.”

The inventor listed on the filing is Benjamin Geza Affleck-Boldt, and the assignee is InterPositive, LLC (i.e. the company Netflix has just acquired).

The patent describes a system that teaches an AI model to watch film footage and understand how the shot was filmed.

Most AI vision systems are trained to recognise objects or actions. They answer questions like:

• Is there a dog in this image?

• Is someone driving a car?

• Is a person walking across the street?

InterPositive is focusing on something slightly different, namely, how the scene was actually shot.

The model learns to recognise cinematic features such as:

• Focal length. This is which lens is being used and how it affects the image's perspective. Short focal lengths create wider views with more background visible, while longer focal lengths compress space and make subjects appear closer to the background.

• Camera movement tracks how the camera changes position during a shot, such as panning, tilting, or moving forward on a dolly. These movements alter how the viewer experiences the scene and can signal emphasis or narrative focus.

• Depth of field refers to how much of the image appears in focus at once. A shallow depth of field keeps the subject sharp while blurring the background, while a deep depth of field keeps both foreground and background clear.

• Camera angle refers to the camera's position relative to the subject. For example, a high angle looks down on a subject, while a low angle looks up, often changing how powerful or vulnerable a character appears.

• Occlusion occurs when one object partially blocks another from the camera’s view. This can reveal information about the spatial arrangement of objects in a scene.

• Parallax describes how objects at different distances appear to move at different speeds as the camera moves. Foreground elements shift more quickly than distant backgrounds, helping convey depth in the image.

In other words, it attempts to convert video footage into cinematography metadata.

The way the model looks at filmed footage mirrors how filmmakers in the real world think about their shots. So InterPositive’s claim that they are looking to use AI to enhance current filmmaking rather than destroy it is borne out in this patent.

I doubt that all but the most ardent anti-AI filmmakers would have a problem with a model that looks at real footage and gives you helpful information about it (assuming the information is correct, of course!).

The model - How does it work?

Their process started with a large collection of video clips, each paired with some form of description or metadata about what is happening in the shot. The system extracts individual frames from these videos and converts them into a format that a machine-learning model can process.

Each frame is then labelled with information describing the cinematic elements present in the shot. This includes the features I already mentioned, such as focal length, camera movement, framing style, and distance from the subject.

Read more

People often complain about a sharp decline in the quality of film writing in the last 10-15 years. So my thinking is, are modern screenplays objectively worse?

I love these types of questions. Partly because it‘s an entry point for research that will help us better understand movies, and partly because it will force us to define terms. In that process of defining something like “worse”, we will learn more about the nature of movies, even before the data reveals the results.

Today I’m going to address one narrow aspect of the question (i.e. writing style).

I studied the dialogue in 64,332 movies released between 1940 and 2022. Each movie’s writing was scored using four metrics, which will allow us to see the change over time.

1. Vocabulary variety (technical term is MTLD) is a measure of lexical diversity, i.e., how long a script can keep introducing new words before it starts repeating itself.

2. Sentence structure, which splits into:

   1. Line verbosity i.e. how many words within a line.

   2. Short lines i.e. how many sentences have only one, two, or three words.

3. Filler words look at how much conversational padding there is, such as “like”, "you know”, “I mean”, etc.

Together, we can use these signals to see whether scripts are becoming simpler. If they are, we would expect to see a narrower vocabulary, shorter speech chunks, more fragments, and possibly more conversational padding.

Now we have our measures, place your bets on what we’ll see…

Signal 1 - Are movies using fewer unique words?

Let’s start with the broadest signal. Vocabulary variety tracks how wide a script’s word range is. Technically, this is measured using Measure of Textual Lexical Diversity which estimates how long a text can continue introducing new words before it starts recycling earlier ones.

In practice, it answers a simple question: how quickly does a script begin repeating itself?

Imagine two versions of the same emotional scene.

1. In one version, a character says something is “astonishing”, then “extraordinary”, then “unprecedented”. The writer keeps reaching for fresh terms.

2. In the other version, everything is “crazy”. The surprise is crazy. The plan is crazy. The villain is crazy. The solution is crazy.

Both versions communicate meaning, but only the former maintains lexical range.

The chart below has blue dots for the average for all movies released in each year, based on an arbitrary base year of 1965. This means we don’t have to be distracted by the exact numbers, but can quickly see whether things are rising, falling, or staying consistent. The orange line is the trend line, to make the trend even easier to spot.

The data is clear - yes, movies are becoming lexically less diverse. Scripts are recycling their vocabulary sooner than they did decades ago.

Dialogue is increasingly drawing from a tighter core set of terms. Characters repeat key words more often, and emotional emphasis leans on reinforcement rather than variation.

When vocabulary range narrows, dialogue becomes more direct, and (arguably) more predictable. The linguistic palette shrinks, even if the story world does not.

So that’s one signal suggesting an increase in script simplicity. Let’s look at the next.

Signal 2 - Are lines of dialogue in movies getting shorter?

In previous research, I have shown that the number of spoken words in movies was at its highest in the 1940s, after which it fell until the 1960s, when it remained fairly static until another decline in the last 1990s. However, since the mid-2000s, it’s been on the rise again.

Connected to this is the inverse trend around silence. The last few decades have seen a significant decline in moments when no one is speaking.

Whilst you may have your own views on whether the lack of silence is a signal for or against the claim that ‘movies are becoming more simplistic’, I don’t think it alone is enough to score this point yet.

We need to dig into how sentences are being constructed. Consider a scene in a movie where one character is confronting another. An older style might run like this:

“If you think for one moment that I am going to allow this arrangement to proceed without objection, then you have fundamentally misunderstood my position.”

That is a single flowing unit of speech which builds, qualifies, and explains.

Whereas, a more modern structure might look like this:

“You’re wrong.”

“This isn’t happening.”

“Not today.”

In those two scenes, the story bears the same, but the delivery is broken into smaller pieces. One would assume that the editing of those two scenes would also differ, with the latter having a fast intercutting rhythm.

Let’s zero in on the length of the lines, so see if characters are expressing complex thoughts or empty short lines.

Line verbosity measures how many words a character speaks before the subtitle breaks. Subtitles divide dialogue into blocks, and each block represents one continuous chunk of speech. Longer blocks usually mean longer thoughts, whereas shorter blocks mean quicker exchanges and more rapid back-and-forth.

And here we see the same pattern as before, i.e., a clear, consistent decline over the past eighty years.

We can go one further on this point and look at the amount of very short lines, i.e. those with only one, two, or three words, such as:

• Wait.

• Why?

• Go.

• Now!

And indeed, we do see a clear rise in these micro lines.

If line verbosity tells us that blocks are shrinking, this measure shows how they are shrinking - ideas are being divided into fragments rather than contained within longer, flowing sentences.

This shift changes the rhythm and reduces the cognitive load on viewers, allowing them to process shorter units more quickly.

So I’m calling that two-zero on the argument in favour of increased script simplicity.

Signal 3 - Are movies full of pointless words?

Finally, we look at empty words or phrases, which take up space but do not drive the story forward or convey information.

Filler words are the small pieces of conversational padding that make speech feel natural. Words and phrases like “like”, “you know”, “I mean”. They rarely carry plot information, but they affect the tone and rhythm.

Compare:

“I don’t agree with you.”

With:

“I mean, I just… I don’t agree with you.”

The latter version feels more conversational, in part because it contains more linguistic padding.

This third signal might be the strongest of the trio, as we saw a stark increase over the period I examined. Characters use more verbal cushioning than they did in earlier films, with dialogue sounding looser, more informal, and closer to everyday speech patterns.

When we pull the three signals together, we can be pretty clear that movie scripts have become:

• Snappier

• More verbose

• But less diverse

• And with more empty phrases

Does this mean they’re worse? That’s outside the realm of data sciences and into interpretation. Thoughts in the comments, please!

Just one more thing…

I suspect some readers are currently thinking:

But Stephen, those change in speech patterns are found in the real world, and so rather than scripts getting worse, this acutally shows that screenwriters are keeping up with society.

And that’s a fair point. Perhaps none of this reflects a decline in writing at all, but instead reflects changes in how people speak in the real world.

It’s fair to say that everyday conversation has become looser, more fragmented, and more filled with verbal padding. So, if screenwriters are mirroring real speech, then what we are measuring is that cultural change, rather than a decline in creative writing.

We need to test if what we’ve tracked so far is down to one or the other of the following:

1. Writing fashion. This sums up the notion that the craft of screenwriting itself has shifted, meaning that modern writers are structuring scenes differently, breaking dialogue down into shorter units and choosing for their characters to speak in faster, more fragmented exchanges. And if you don’t like that, you might conclude that ‘films are getting more simplistic’.

2. Real-life reflection. Writers reflect the world as it should be for the settings of their movies, and as society has become more relaxed, cinema has followed suit, with most set in the present day. If this is true, then scripts are not getting simpler, but we are getting simpler, and movies are just a mirror.

How do we test which is true? We look for movies set in the same time period but written at different points in the past eighty years.

I identified two cohorts of movies from the wider database:

• Contemporaneous movies are those which were set in the same decade in which they were released.

• Movies set in the 1800s, as this was the largest time period outside of the 20th and 21st centuries. (More on that here: What is the most popular time period for movies?)

If these changes are about writing fashion, then we should see the same drift we saw earlier.

But if the writers are accurately reflecting the time period their stories are set in, then there should be no change over the past eight years.

So what do we see?

For our first two structural signals, the cohorts move together.

Dialogue lines are getting shorter in both groups.

And the number of very short lines is increasing in both groups.

That suggests that at least part of what we’re seeing is not about characters sounding more “modern”. It looks more like a shift in writing style, with scenes broken into smaller conversational beats, exchanges tighter, and thoughts divided into fragments more often.

That structural compression is happening regardless of whether the story is set in 2022 or 1822. So, some of this change is about how films are written and edited, not about what period they depict.

But when we look at language complexity, the picture changes.

Films set in the 1800s do not show the same decline in vocabulary variety that we see in contemporaneous titles. There is still some downward movement, but it is noticeably flatter.

The clearest divergence appears in filler words. Over the period I studied, filler words increased markedly in films set in the present day. In films set in the 1800s, they slightly declined.

Writers of period films appear to resist some aspects of modern speech patterns. Excessive filler words would feel obviously out of place in an 1800s setting. In previous research, I have shown that audiences punish films they regard as using anachronistic language choices, which could explain this heightened awareness (or pressure from producers).

So what have we learned?

Across the past eighty years, movie dialogue has become structurally tighter, with shorter lines, very brief exchanges becoming more common, and scenes are broken into smaller conversational beats.

This shift appears in both contemporary and historical films, suggesting it reflects changes in writing fashion and pacing.

At the same time, vocabulary diversity has declined, and filler words have increased, but mainly in films set in the present day. Period films do not follow that trend to the same extent. That implies that some of what we are measuring reflects contemporary speech patterns, not a universal drift across all storytelling.

In short, we are seeing two forces at work: a broad structural compression in how dialogue is constructed, and a setting-dependent shift towards more conversational, padded language in modern-set stories.

Movies may well be becoming structurally tighter and more conversational. But whether that counts as “simpler” depends on whether you see those changes as loss, adaptation, or evolution.

What do you think?

Notes

This research examined the English-language subtitle files of 64,332 movies released between 1965 and 2022, inclusive.

Each film was scored on four subtitle-derived language signals.

• Vocabulary variety was measured using MTLD (Measure of Textual Lexical Diversity), where lower values indicate faster repetition of the same word set.

• Line verbosity was measured as the mean number of words per subtitle block, with lower values indicating shorter speech chunks.

• Structural fragmentation was measured as the share of subtitle lines containing one to three words, where higher values indicate more micro-lines.

• Conversational padding was measured as filler-word frequency per 1,000 words, with higher values indicating looser, more padded dialogue.

After metric extraction at the film level, change over time was estimated from yearly medians rather than yearly means to reduce the influence of outliers. Trend direction was then modelled with weighted linear fits across release years, with yearly film counts used as weights so denser years contribute more than sparse years.

Time settings were initially determined based on explicit mentions in film synopses, which could reference a specific year, decade, century, or a general time period such as “medieval.” Films with no clear indication of their time period were assumed to be set in the decade of their release.

Subtitles are not perfect proxies for scripts, but they are close enough to capture structure, vocabulary, rhythm, and repetition. One thing to note is that subtitle formatting conventions could have changed over time, which could affect some of what we saw. I’m not too worried about this, as it alone is unlikely to explain the divergence between contemporaneous films and those set in the 1800s. If subtitle fashion were the primary driver, we would expect both cohorts to move in near-identical ways across all four signals, but they do not. So while subtitling norms may account for part of the structural shift, they are unlikely to be the main explanation for the differences we see between story settings.

In the past, I have looked at other aspects of the “Are movies getting worse?” debate, including:

• Are movies becoming more derivative? looking at the script’s original source material.

• Has the way movies use dialogue changed? tracking the amount of dialogue and the gaps between characters’ speech.

• How often do lead characters die in movies? and What percentage of movies have a plot twist? address the changing nature of the stories being told on screen.

• Are film critics losing sync with audiences? and How one-star and ten-star votes break IMDb’s user score both look at how people rate movies at movie ratings.

It's frequently claimed that NASA uses the 1998 blockbuster Armageddon as a management training tool, asking its boffins to list all the factual errors they can find in its bombastic 150-minute runtime (reportedly up to 168 errors by some counts).

Hollywood is largely unbothered by this type of criticism. On the DVD commentary track for Armageddon, Ben Affleck asks the film’s director, Michael Bay, why it was “easier to train oil drillers [to become astronauts] than astronauts to become oil drillers”, to which Bay responds, “Shut the fuck up”.

Nonethe less, I do care about errros. Alot.

I collected a dataset of over 200,000 errors reported by the public across almost 12,000 movies and classified them. (There are more details on my methodology in the Notes section at the end).

Let’s see just how wrong movies are…

What’s an error?

Before we get into the details of specific error types, it helps to split the data into two broad groups:

1. Story errors. These are errors created at the script stages, such as plot holes, incorrect facts, how things work, geographic impossibilities, inaccurately detailed historical events, character logic failures, factual mistakes, or anachronisms.

2. Filmmaking errors are those caused by the physical process of making the film, taking place mostly (but not exclusively) on-set, such as continuity slips, crew in frame, equipment visible, audio issues, etc.

In most genres, filmmaking mistakes are more common than story mistakes. They account for roughly two-thirds of mistakes in musicals (67.4% of errors are filmmaking-based), horror (67.2%), fantasy (66.7%), and comedy (66.7%).

In these genres, viewers are more likely to point out continuity issues or production slips than problems with the underlying logic or research. My theory is that this is because these genres feature more fantastical stories or characters, so facts count for far less.

At the other end of the spectrum, we have more grounded genres, such as war (43.8%), biography (32.8%) and history (31.8%). In historical films, more than two-thirds of reported mistakes concern the story or the facts rather than filmmaking techniques.

So when a genre trades on realism or real-world knowledge, the more we as the audience focus on research and internal logic. In genres built around spectacle or heightened reality, attention shifts towards visible production errors.

How many films make factual errors?

Factual mistakes are the most common type of error, even in these least scrutinised genres. 44.1% of music films had at least one reported factual mistake, followed by romance (49.3%) and musicals (50.2%).

At the high end, there are two different factors at play:

1. In historical (71.2%) and biographical (68.1%) films, the pressure comes from documented reality. Over three-quarters of war films have reported factual errors. We compare what we see on screen with what we already know (or think we know) about events, dates and people.

2. In action (71.9%) and sci-fi (72.3%), the pressure comes from systems. If a film builds a world around physics, engineering, or technology, we are on the lookout to see whether those systems behave consistently.

Part of my work for Guinness World Records is determining the best criteria to answer subjective questions. So if one were to ask, “What is the wrongest film?” then my first stab at an answer might be The Martian.

The Martian has the highest number of reported story errors per minute of screen time.

Among the movie’s alleged inaccuracies are:

• Mars itself is wrong: storms with enough force to tip hardware, plus Earth-like sunsets/Sun size and near-Earth gravity.

• Space physics is fudged: objects float during acceleration, and astronauts change direction in microgravity without pushing off anything.

• Comms and maths are dramatised: near real-time messaging despite 4–24 minute one-way delays, and orbital mechanics treated as needing days/supercomputers.

• Life support is off: suit pressure/oxygen readings don’t add up, decompression protocol is ignored, and the “Hab fabric” patch behaves impossibly under pressure.

• Chemistry/engineering shortcuts: hydrazine reaction explanation doesn’t fit the scene; mismatched voltages would fry Pathfinder; display units/labels are wrong.

• Farming is overstated: perchlorate-toxic soil untreated, sunlight/water assumptions optimistic, and potato planting/seed quantities don’t match the crop shown.

• The list goes on…

How many films feature continuity errors?

Moving on from story errors, we should now look at filmmaking errors. These are the ones that reveal the artifice of movie-making and can take us out of the story and its world.

The most common of these were continuity errors, in which something changes between shots meant to take place at the same moment. Popular examples include:

• In Pulp Fiction, bullet holes are visible in the apartment wall behind Jules and Vincent before Marvin fires the shot that supposedly creates them.

• In Harry Potter and the Philosopher’s Stone, the level of drink in Harry’s goblet shifts between cuts during the Great Hall scenes, despite no one touching it.

• In The Dark Knight Rises, Batman’s cowl switches from damaged to intact and back again during the same street fight sequence.

• In Pretty Woman, Julia Roberts’ croissant turns into a pancake between shots during the breakfast scene, then back again.

When we drill down to the objects that cause the most continuity issues, we get a sense of the scale of the challenge filmmakers face when trying to shoot a coherent 90-minute story over multiple months in a semi-random order.

Cars and trucks are at the top, containing the obvious challenge of trying to get traffic to do the same thing for multiple takes, or to accurately predict how a yet-to-happen crash will affect a vehicle you need to shoot in now.

Small handheld items don’t face this problem, but can be easily changed on the spot and feature in a close-up. Hence, why drinks and weapons are the next biggest challenges on-screen.

Which areas cause the most anachronisms?

A close cousin of both factual and continuity errors is anachronisms. These occur when something in a film does not belong to the time period in which the story is set.

These include:

• Oppenheimer features 50-star flags at Strauss’s 1959 hearing.

• Sully includes an orange “Empire Gold” New York souvenir plate in a 2009 airport scene.

• Men in Black 3 uses the wrong Spanish flag for its 1969 setting.

• Braveheart shows people wearing baseball caps and driving modern white vans. (I wish I were joking).

The data here again reflects the complexity of accurately depicting vehicles in a movie, with almost 6% of movies featuring some kind of wrong or anachronistic vehicle on screen.

The next most common areas are brands (i.e., the product was unavailable during the period or at the location where the movie is set), architecture, and technology.

Where do films most often break the laws of physics?

Even if the filmmakers get the right object on screen, audiences notice when it behaves incorrectly.

I searched all the factual references related to physics and examined what dominated. The most common such complaints centre on classical mechanics, such as forces, motion, inertia, falls, and impacts - pretty much the core elements of an action movie.

Next, we have materials science (15.2%) and fracture mechanics (7.3%), i.e. what bends, what snaps, what shatters, and what somehow survives a hit.

There’s too much in this topic for one article, so I shall return to it next week to look at plot holes! Ensure you’re subscribed to get the info first.

Subscribe now

Notes

This research examined over 200,000 issues or errors across 11,842 movies released between 1930 and 2025. My initial sources were IMDb, Wikipedia, and other movie sites.

Given that these are errors reported by the public, there is a strong bias towards larger, more closely scrutinised movies. We can’t track every frame of every movie and generate new data, partly for time reasons (I’ve got other stuff on, you know), but also because much of this area remains subjective. One person’s error is another person’s legitimate movie detail. For example:

• Is Rose's refusal to share the ample floating door space in Titanic a plot hole or just selfish character behaviour from someone who has gone off her holiday bit of rough?

• In the Star Wars universe, is the stormtroopers’ inability to hit a stationary target a story error, or a stark warning of what happens when military costumiers prioritise fashion over function?

• Maybe in James Cameron’s 1997 epic, Titanic, the disappearing handbag in the Andrews conversation is not a continuity error but rather error-free evidence that Rose is a world-class thief, trained in high-society sleight of hand since birth.

• Or in Titanic, is Rose’s seeming inability to calculate basic buoyancy physics a character error, or a damningly accurate illustration of the education level of snobs of the time?

We just can’t be sure. Oh, wait, we can.

An episode of A Knight of the Seven Kingdoms briefly climbed into the same rarefied space as Breaking Bad’s “Ozymandias”. Within hours, users were mobilising. Some piled in with 10-star votes to defend the newcomer. Others countered with 1-star votes to protect an incumbent.

None of this is new. IMDb has long had to contend with coordinated vote-stuffing and defend against the claim that its official “IMDb Rating” is biased and/or broken.

But how much do one- and ten-star reviews actually affect the final IMDb score? I want to take you on a journey behind the numbers to provide context for the current (and future) arguments.

To do this, I looked at over 85 million votes cast on IMDb across 73,000 movies.

What is an “IMDb Rating” anyway?

On the surface, it is a single number between 1 and 10, rounded to a single decimal place. Most people use it as a quick, effective way to say whether a film is any good. For example:

• Terminator 2: Judgment Day has a score of 8.6 out of ten, putting it in the top 3.5% of movies on IMDb.

• Terminator 3: Rise of the Machines has a score of 6.3, putting it in the bottom 52.2%.

We all like a single specific score. The number looks precise, it implies consensus, and is easy to quote. Also, it’s useful. A quick glance at a movie’s IMDb rating is often enough to nudge us towards or away from investing time in a movie.

And yet, as soon as you put even a modicum of scrutiny on this number, you realise that it is deeply flawed. Among the problems with it are:

1. It pretends we agree. A single number implies that there is an objectively ‘correct’ measure of a film’s quality, hiding any sense of polarisation or disagreement.

2. It ignores the extremes of love and hate. Extreme and intense feelings matter. We all have movies we will defend to our last breath, and others we couldn’t be paid to watch again. A single rating doesn’t reflect the levels of extreme love that 10 out of 10 votes reflect, not those scoring it just 1. (More on this in my piece from a few weeks ago, What makes people never want to rewatch a great film?).

3. What is it even measuring?! Some see the IMDb score as a signal of a film’s inherent artistic quality, as in “This is a good film”. It can also be used as a reflection of a film’s appeal, as in “This is a popular film”. Another reading could be of a movie that people want to see more of, as in “I support this kind of movie”.

4. It doesn’t tell us what we actually want to know. If you’re picking a movie to watch, there are so many factors that go into your choice that a single, averaged number from the wider public is close to useless. I would guess that the extent to which a film is entertaining would be greater than its artistic merit. Schindler’s List is the 6th-highest-rated film on IMDb at 9.0, but few would casually add it to their list when looking for a movie to pass the time. (See Which types of movies make money despite receiving bad reviews from critics?).

5. It ignores who is voting. People are not in a single homogeneous group, and so a single score ignores age, gender, location, cinephilia, etc. Worse than that, one group may dominate: when I last crunched the data, 81.7% of IMDb votes were cast by men, and 45.0% by people aged 30-44.

   

6. Early voting is not representative. People who see a movie early behave differently from those who see it late, but a single average cannot distinguish between them. This is especially important if the score is used to determine audience behaviour in the crucial opening week of a movie's release.

7. It can be gamed. Vote stuffing to increase or decrease a movie’s score is rife. More on that in a moment.

Both the films below (Companion and Vettaiyan) have an official IMDb score of 6.9, but show very different voting patterns.

An uneven distribution of votes is not uncommon.

This is the world in which IMDb has operated when it’s calculating its final official IMDb Score. The raw data it’s dealing with (i.e., the votes cast) are frequently extreme, reflecting a desire to move the needle rather than passively reflecting the voter’s measured opinion.

There are so many films with voting patterns that simply cannot reflect what viewers think of the movie's quality.

How many IMDb scores are skewed?

There are three ways in which a film could be affected by vote stuffing:

1. Push-down, where unusually heavy low-end voting (especially 1-star) is dragging down the average.

2. Push-up, where unusually heavy high-end voting (especially 10-star) is bringing up the average.

3. Collision voting, in which both ends are elevated at once (lots of 1s and lots of 10s), results in the score moving in either direction, depending on edge strength.

To get a sense of how many films experience these forces, I classified each film by comparing its 1-star and 10-star vote shares against what is normal for films with similar average scores in the same country.

• If neither extreme is unusually high, it goes into normal

• If only 1-star is unusually high, it is push-down.

• If only 10-star is unusually high, it is push-up

• And if both extremes are unusually high at the same time, it is classified as collision voting.

This reveals that over 10% of films on IMDb are experiencing extreme edge pressures.

Is it getting worse?

In a word, yes.

Below is a time series showing the rate of extreme voting behaviour. It reveals that since the early 2010s, the number of 10-star reviews has increased massively. One-star voters saw a slight increase, followed by a return to the longer-term average in the most recent years.

Eagle-eyed readers will have already spotted two anomalous years in the graph above. The number of ten-star votes in 1994 and 2008 is strangely high. And this neatly highlights another issue with movie ratings - outliers.

The spike in those two years is almost entirely down to just two movies - The Shawshank Redemption (1994) and The Dark Knight (2008).

54.9% of all the votes cast for The Shawshank Redemption were ten stars. Given that it’s also the site’s most rated film (3.2 million votes to date), this is enough to disrupt the whole year’s ratings. Similarly, 45.6% of The Dark Knight’s 3.1 million votes were full marks.

If we remove just those two titles, the chart reads a little clearer.

Where in the world is this happening?

These push-pull effects are not distributed equally.

IMDb now shows country-level vote breakdowns, showing the vote counts for the five highest-voting countries for each movie.

The differences between voters in each country are likely to be twofold:

1. Cultural differences around what it means to be a fan. In some countries, vote-stuffing may be seen as a proper way to support your favourite movies or star, while in others it may be frowned upon, no matter the fervour of your fandom.

2. Movies of national relevance. Cinema has always been political (no matter what Wim Wenders says), and that’s never clearer than when a film tells a story associated with a nation’s identity. These are more likely to whip up pride and anger, leading to strategic voting.

Fans in India, Bangladesh, and Turkey are most likely to leave one- or ten-star reviews online. Spanish, Swedish, and American fans are the least (from the countries with enough voting data).

What types of films attract the most abnormal voting?

Genres attract different kinds of audience behaviour.

Perhaps appropriately, war films receive the most abnormal voting behaviours. I put this down to their high propensity to be stories of politics, identity and national history.

It seems reasonable to assume that most war films make one side out to be more virtuous than another, and so the number of people who feel offended is likely to be higher than the number of people who feel offended by romance films. (I guess one advantage of having your victim be uncaring, shitty boyfriends over a particular national group is that the very people you are slating don’t know or care that it’s them being attacked).

This highlights a point I made earlier, that a 7.2 average can hide wildly different realities. For a rom-com, it might mean most people are genuinely settled around a 7. For a war film, the same 7.2 could reflect a split crowd, with a chunk handing out 10s and a larger chunk voting it down to 1.

Who’s doing this?

So far, I have loosely referred to the people providing the edge votes as “fans”, the truth is likely to be messier than that. They could include:

• Supporters of a movie, star, story, or theme. These are people who want their interests to be better regarded and/or to receive the recognition they so obviously deserve.

• Haters of people, ideas, or trends. Lots of movies get systematically downvoted because they take a position on politics, race, or religion. Or simply because it’s something young women like.

• Representatives of the people involved. I’m sure there must be people who represent a tricky client and have twigged that when their client’s projects' IMDb scores go up, the client becomes more tolerable. Plus, a growing recognition for their work may also lead to commanding higher fees or better terms in future negotiations.

• The stars themselves. Heavens forbid that I suggest, for even a moment, that actors may care one jot about what people think of them or their movies. Equally, I know that all filmmakers never read reviews, as they are constantly telling me so.

Given the large number of people who may be motivated to affect an IMDb score, we cannot say that just because a certain individual repeatedly benefits from vote stuffing, they are the cause of it. It’s very possible they don’t even know it’s happening.

So as a public service to those unwitting winners, I looked into which stars and filmmakers seem to repeatedly have their projects targeted for absurdly large numbers of 10-star reviews, when compared to films with similar overall scores and vote volumes.

Is this a problem with kids today?

The demographic data on IMDb used to be somewhat limited, and today it’s nonexistent. But using older data, and some clever maths, we can estimate who’s behind the extreme behaviour.

I mapped each demographic group’s average rating onto the actual vote shape of the films they rated. From that, I could infer how much more or less likely that group is to use 1-star or 10-star relative to a matched baseline.

The chart below shows the results, albeit in a slightly unintuitive format. Here’s how to read it:

• Each row is a demographic group, using IMDb’s classification (note that in this case, ‘old’ means everyone between 45 years old and death).

• The red dots denote that group’s relative tendency toward 1-star voting and the green dots show their relative tendency toward 10-star voting.

• The vertical dashed line at 1.0 means “no difference from baseline”.

• So if a dot sits to the right of 1.0 then that group uses that extreme more than expected. If a dot sits to the left of 1.0 then they use that kind of extreme voting less than expected.

• The farther away from 1.0, the stronger the tilt.

What we discover is that young people love things too much, and old people hate things too much.

So how does IMDb respond to this?

I hope you have grasped by now just how hard IMDb’s job is. They are being bombarded by millions of targeted votes, designed to maintain a score that many people take seriously and which is a key part of IMDb’s brand and credibility. I do not envy them nor the task they face.

IMDb does not have a creative agenda. It would not work to adjust scores on a case-by-case basis, partly because they are trying to be apolitical, but also because dealing with the sheer number of people who believe they are suffering from “unfair” voting would be unworkable.

They need a single system that applies equally to all films.

(Well, the voting on This Is Spinal Tap claims to go up to eleven, but (a) that’s the only case I know of; (b) it doesn’t really, it just says that on the film’s homepage; and (c) to be fair, that’s a pretty good joke).

To tackle this, the official IMDb rating is not an average of all the votes. Instead, they use a secret algorithm that seeks to account for extreme voting while acknowledging that some movies do actually deserve large numbers of one- or ten-star votes, as we've seen with the Batmen of Justice League and The Dark Knight, respectively.

How does this algorithm work? It’s a proprietary secret. And it should be, as sharing the formula can only help those looking to manipulate the system.

My best guess is that it uses;

• The arithmetic mean from the full histogram,

• The total vote count,

• Signals about the vote distribution shape, especially extreme buckets,

• Interactions between 1-star and 10-star pressure, and

• Possibly trust-weighting based on user data, such as account age, voting history, behavioural consistency, detection of coordinated bursts and other anti-bot measures.

Below are a couple of extreme examples, both with an unweighted mean of around 7.3, both of which have much lower scores than their mean, but with different IMDb scores due to different voting patterns.

In almost four out of five movies, the official IMDb score is lower than the arithmetic mean. I presume this is because every project has at least some people motivated to stuff 10-star votes - i.e. the cast and crew!

The Goat Life is one of the rare films where the official score (7.1) is higher than the unweighted mean (5.7).

IMDb compared with Letterboxd

IMDb is also operating in a more combative environment than other sites. Because it's the biggest and ‘site of record’ for many people, it attracts the most attention from those wanting to make a point, promote a film they like, or punish a film or person they don’t.

A quick check of comparable scores on the Cinephile database Letterboxd shows that IMDb has roughly twice as many extreme votes.

Overall, the two sites agree quite a bit, with a correlation of 0.80 for films with over 1,000 votes on both platforms.

In defence of the IMDb user score

I’ll be honest and say I think the IMDb system does a pretty good job, given the vast influx of messy data and the need for a single universal solution.

Here’s the real truth - this is not a problem with IMDb. It’s a problem with us. By us, I mean both:

• The people who vote. If the invention of the internet has proved anything, it’s that when people can hide behind a keyboard, they act like trash. In this article, I have been rude about actors, agents, boyfriends, Batmen, and suggested that the next life milestone a 45-year-old has to look forward to is death. I wouldn’t do any of this in person. When you add to the mix that people feel very passionate about movies they love, this leads to a lot of mean behaviour.

• The people who use IMDb score. As I said at the top, it is foolish to think that any score can flatten all the factors we've seen and accurately reflect a piece of art with a single number. IMDb does its best to provide a quick guide, but at the end of the day, we only have ourselves to blame if we think the score matters, and if we are offended by it.

The IMDb user score is not wrong - we're just using it wrong.

Epilogue

It’s easy for this to become purely about maths, but there is a human side to this topic.

When a movie is targeted for mass negative commentary and voting, real people will be hurt. The negative effects fall on directors, producers, cast and crew. It shapes perception, affects confidence, influences press coverage, and becomes part of the public record.

One such person is the Iranian-born British artist, director, and screenwriter Tina Gharavi. Gharavi is the director of the four-part miniseries Queen Cleopatra starring Jada Pinkett Smith, Adele James, and Craig Russell.

The series currently has an implausibly low IMDb score of 1.2 out of 10. One-star reviews account for 93.7% of the 88,000 votes cast. (If we drill down to just voters in Egypt, then it’s even worse at over 97% one-star votes).

That’s half the rating the classic stinker Battlefield Earth got from roughly the same number of votes cast. It’s also at odds with the critics' view, with the project scoring 45 out of 100 on Metacritic.

I spoke to Tina about this, and she explained the effect this has had:

The reviews and ratings were part of a coordinated and politically motivated hate campaign, largely in response to the depiction of a Black Cleopatra. It was never truly about the quality of the production. I welcome rigorous debate about history, representation, and creative interpretation - that’s part of the job. But this did not feel like debate. It felt like a personal and racialised attack.

Ironically, the campaign generated enormous publicity and drove viewership. The series reached number four globally on Netflix and number one in several territories - probably not the outcome its organisers intended. What is regrettable, however, is that malicious and racially motivated voting can remain embedded on major platforms long after the initial outrage cycle has passed.

It’s not hard to imagine how Tina feels. She is experiencing a sustained and targeted attack on her and her artistic work, from people who are unlikely to have watched it.

When it first happened, Tina reached out to IMDb to note that the voting was anomalous and asked whether they agreed. IMDb’s reply included the following:

Please rest assured that the rating for the titles in your filmography is correct, as generated using the same method employed to calculate the rating for all titles listed on IMDb.com, which includes numerous safeguards to detect and minimize the effects of attempts to skew the results. Please also note that some of those titles have received a very small number of votes, which makes them susceptible to a lot of variability.

The hard thing to acknowledge here is that IMDb isn’t wrong. Assuming they are successfully removing suspicious votes from bots, the histogram does reflect how 88,000 people voted.

One could argue that they are platforming hateful views, as expressed through votes, but that’s not the same as being at fault.

All ratings are inherently subjective, and we can’t say that one subjective reason is more legitimate than another. Both because that’s the nature of subjective opinions, and because we have no way of knowing what caused them to vote that way.

But where does that leave people like Tina?

A few years on, I’ve learned to process it differently - but that doesn’t mean it leaves no mark. When a project is reduced to a number that has clearly been distorted, it enters the public record in a way that can affect perception, opportunity, and confidence - not just for me, but for the hundreds of people who worked on it.

I understand the philosophical argument that all votes are subjective. But when tens of thousands of one-star votes appear before most people have even had the chance to watch the work, it’s hard to see that as organic audience response.

What concerns me most is the precedent it sets - particularly for work that challenges dominant narratives or centres historically marginalised perspectives. If coordinated outrage can permanently brand a project with an artificially low score, that creates a chilling effect. Artists should be accountable for their work, but platforms also have a responsibility to examine how their systems can be weaponised.

Today’s research has reminded me that so often the industry reduces things to data and forgets that there are real people at the sharp end of these debates.

I wanted to leave the final word to Tina:

When a platform allows coordinated, bad-faith campaigns to permanently distort a work’s public record under the banner of “neutrality,” it is no longer a passive host of opinion but an active amplifier of harm - and with that amplification comes responsibility.

Notes

The research today focuses on movies, not TV or other types of content.

All the demographic data was up to the end of 2015. It's been a while since I last looked at the age and gender of IMDb voters, and since then, IMDb has stopped providing that data, instead showing combined scores split by the top five countries for each movie. It’s totally possible that the demographics of IMDb’s user base has radically changed over the past decade, although that feels unlikely.

When referring to years in the charts, this relates to the movie’s release, not the year the votes were cast. So it’s reasonable to assume that most of the votes cast for movies since around 2000 were within the first year or so of release, but certainly for anything before around 1995, it’s voting with the knowledge of history on our side.

To define the “push-down / push-up / conflict” I measured how unusual a movie’s one-star share and ten-star shares were relative to peers in the same country and similar score band. This creates standardised anomaly scores (z-scores). Using a threshold of z >= 2:

• Push-down = unusually high one-star, but not ten-star

• Push-up = unusually high ten-star, but not one-star

• Conflict = both unusually high

• Normal = neither unusually high

Part of the reason I was able to whip up a long article on short notice is that I have been studying ratings, including those on IMDb, in the background for a long time. Ten years ago, I teamed up with a professor of Astrostatistics to identify the creators and actors most associated with projects that received an unusually high number of up-push votes (as I did again above).

We created a list of 100 top candidates, and I emailed them all, offering them a chance to chat off the record. Perhaps to no one’s surprise, I got exactly zero replies.

If the video has made you want to explore the topic a little deeper, here are some past research pieces I’ve conducted:

• Is ‘Go Woke, Go Broke’ true for movies?
  My analysis of 10,000+ films and millions of audience reviews to test whether “woke” themes actually hurt box office performance. I break down where the slogan holds up, where it doesn’t, and how genre, budget and execution shape the outcome.

• Measuring if movies are too woke
  A full methodological breakdown of how I quantified “woke-adjacent” themes in film reviews, built thematic-similarity models, and tracked changes in cultural discourse around movies over time.

I’ll be uploading new video essays every couple of weeks throughout 2026 (if YouTube will let me), so make sure you’re subscribed on YouTube to get notified when another drops (also, subscribing will really help me out with the Algorithm Gods, so it would mean a lot to me - thank you!).

https://www.youtube.com/@TheFilmDataScientist

Here is the previous video:

And to help explain the channel, I also created a short channel trailer:

A major reason I attend these events is to listen to what people talk about, what they ignore, and how they frame the industry. The key things I picked up this year were:

1. Risk mitigation has become the industry’s default setting

2. The market keeps getting shorter

3. AI is fully embedded in the market (but not in people’s hearts)

4. The movies were less stary-y and more art-y

5. Stars tried to be apolitical at the most political of festivals

6. Russia’s changing role (again)

7. TIFF’s market steps into the limelight

Mad festival stuff

Before we get into the serious topics, I thought I’d open with something lighter. Long-time readers will know of my ever-growing list of ‘mad things I’ve genuinely overheard at film festivals’. This year’s EFM gave us these beauties to add to the list:

• “The only studio contact I have is my dealer. They supply some execs”.
  Response: “Do you think we can get them to read the script?”

• On three separate occasions, I heard the phrase “Nothing good happens at a film festival”. I’m thinking of putting on a t-shirt. Or better still, I could follow the Berlinale‘s approach and put it on a hoodie for €53.

• “It’s written by two women, so you know it’s safe from that point of view” (This was a meeting of four men in preppy sweaters and chinos).

• A guy who was being asked a series of very valid questions about his project, to which he replied to all with “I wish I knew” in a tone that implied he thought that was a full, satisfying answer. He was just so keen to be funded and find out the answers to these great questions.

• “We’ve got a lot of women in peril films” Looking at the producers in question, it’s worryingly unclear if they were talking about on or off screen.

Ok, that’s enough for that. Let’s look at each of the key things I noticed in detail.

1. Risk mitigation has become the industry’s default setting

My main takeaway from this year was risk mitigation. A lot of what was being said was rooted in actions people were taking in response to perceived risks.

That shows up first in how organisations position themselves. The EFM’s 2026 messaging leaned into “new strategic partnerships” and “cross-sector collaboration” to address an industry that feels more fragile and interdependent than it did a few years ago.

I heard this a lot, with increased chatter around bridge-building between festivals, markets, funding bodies and companies, because access, leverage and exposure feel harder to control than they used to.

It also shows up in deal structure and rights thinking. The Ryan Coogler “Sinners” deal became a reference point in conversations about ownership, reversion and long-tail value. Not because everyone can get those terms, but because it puts “who ends up owning this in 10–25 years?” back into the mainstream bargaining vocabulary.

Finally, it shows up in how projects are financed and protected. Tax credits and incentive regimes remain the most relied-upon form of risk reduction, and jurisdictions keep tweaking them because production is mobile and capital is cautious. In practice, this pushes producers towards structures that can be collateralised and audited, and away from anything that relies on optimism.

An adjacent neighbour of this was the sharp rise in casual nihilism. It seemed that at every turn and in every conversation, there was a weary acknowledgement that everything is kinda broken, but that we should keep showing up to meetings nonetheless. “But what can you do…?’ was how most conversations resolved.

Maybe THIS should be my hoodie slogan?

2. The market keeps getting shorter

The very first edition of Screen’s dailies (published on Thursday) said:

Today may be the first official day of the EFM, but many sales executives have been in town since Monday… Some buyers are planning to leave Berlin by the weekend.

This matches what I saw on the ground, namely concentrated early meetings, tighter schedules, and less mid-week drift. This is not necessarily a bad thing, nor does it denote a weak marketplace (in last week’s piece, I showed how attendance is still growing year on year).

This change is down to an efficiency drive. The time-consuming bits can be handled via email, Zoom, and screeners, leaving just the human interaction for in-person. Plus, we’re seeing the continued deprioritisation of the party/social side, as seen more often in Cannes and pre-pandemic Berlin.

Markets are no longer assumed to require full-time in-person attendance, as deals can continue remotely, meaning the all-important in-person window can be compressed.

3. AI is fully embedded in the market (but not in people’s hearts)

As you might expect, AI was ever-present, making its way into workshops, ways of working, and real projects on sale in the market.

Events. Among the EFM 2026 events focused on AI integration were:

• “From Hype to Help: Real-World AI Use Cases Worth Your Time”

• “Pitch Perfect: Producers Showcase Projects Analysed by AI” (Yes, the era of AI focus group approaches).

• Not all were presenting a positive framing of AI (another way of putting that is… not all were paid for by companies that sell AI products), such as “Rethinking Producing Through Collective Power to Sharper decision-making in the AI era”

Expectations. Across the board, AI is framed as part of how films get financed and packaged, rather than as some speculative future topic. More than a few filmmakers I spoke to said they were being expected to bring AI-generated videos to pitch meetings, even for non-genre projects. This mirrors a survey released this weekend that found that half of the 254 professional Indian screenwriters surveyed said “producers now expect faster turnaround times because they assume writers will use AI”.

Exploitation. On the sales side, a number of AI-led projects were on show:

• Odin’s Eye Entertainment offered an AI-generated documentary Post Truth for international sales.

• The Hybrid Creators pitched RebelDogs featuring “hybrid actors” that buyers could overlay with local celebrity faces using AI.

• As well as possibly the first AI-generated film to qualify for the Oscars, The Great Reset was being sold at EFM.

Read more

The honest answer is just time.

But over the past few months I have been quietly beavering away, building up a bank of videos which bring some of my past data studies to life in a fun way.

I’ll be uploading new video essays every couple of weeks throughout 2026 so make sure you’re subscribed on YouTube to get notified when another drops (also, subscribing will really help me out with the Algorithm Gods, so it would mean a lot to me - thank you!).

https://www.youtube.com/@TheFilmDataScientist

To help explain the channel I also created a short channel trailer:

(If you’re going to be in Berlin too, reach out, and we can grab a coffee.

Berlin tends to be more compact and siloed than Cannes, but just as useful for understanding the market. You can read my summary of last year’s EFM here- Six big takeaways from the Berlin European Film Market 2025.

To prepare for meine Reise zur Berlinale, I have gone through the programme with a fine-tooth comb to get a sense of what awaits us.

I’ll go through it in detail later in the article, but here are the key themes as I see them:

1. The biggest conversation is about how to get through the next few years.

2. AI has become everyday work.

3. The franchise dream is being handled more carefully.

4. Audiences are part of the plan from day one.

5. Technology as tools rather than magic.

6. Environmental impact is something you now have to account for.

7. Fair work and representation are built into the conversation.

8. Rules and regulations are shaping what gets made.

If you’re attending, you may enjoy this year’s EFM, then you may enjoy my breakdown of tips from 237 film industry professionals - Tips for attending the European Film Market

Let’s start by looking at what the EFM has been about in the recent past, before going through those themes in more detail.

Two decades of EFM sessions

To put the 2026 programme in the right historical context, I began by reviewing the past 20 years of EFM sessions. It was a fascinating journey that showed perennial problems as true today as they were then, alongside things that seem like distant memories (3D, VR, and NFTs, oh my!).

But in those new buzzwords and technologies, the same themes keep popping up. Virtual reality echoes earlier optimism around 3D cinema. NFTs replay familiar arguments about ownership and scarcity that are at the heart of most value extraction. Perhaps direct-to-fan platforms are just the latest version of many a producer’s long-running ambition to bypass distributors.

After reading a whole bunch of old EFM programmes, Internet Archive snapshots and contemporaneous press reports, here’s what 20 years of EFM sessions felt like to me:

• Late 2000s to early 2010s - An industry trying to steady itself. In the late 2000s and early 2010s, the programme reflects a sector facing economic shocks and structural fragility. Financing, public support, and market confidence dominate the agenda with headlines like “EFM Industry Debates to focus on crisis“ (2009) and “Europe’s Film Industry in Crisis” (2013), squarely placing the focus on stability and survival.

• Early 2010s - Digital tools as a source of hope. As the decade turns, attention shifts toward technology as a possible corrective. Online distribution, audience access, and new financing routes are framed as ways to bypass weakening intermediaries. Sessions like “Crowd Funding – How to Harness the Power of the Online Audience” (2012), and “Buying and Selling World Cinema: Reaching Audiences in the New World” (2012) capture this period of optimism.

• Early to mid-2010s - Technology moments that dominate the halls. Certain years are shaped by the rise of a single technology that briefly defines the conversation. The 3D cycle is unusually visible, from “3D – Fab or Fad?” (2011) to “3D After the Hype” (2013), just a few years later. Comparable arcs later appear around VR and blockchain, each receiving concentrated attention before settling into narrower use.

• Mid-2010s - More responsibility pushed onto producers. As traditional gatekeepers lose some authority, producers are encouraged to take greater control over financing, rights, and distribution. Independence is increasingly framed as a practical requirement rather than a creative preference. Panels including “Producers as Self-Distributors” (2015) and “Producers as Entrepreneurs“ (2016) reflect this shift in responsibility.

• Mid-2010s to late 2010s - The rise of series thinking. From the mid-2010s onward, television and series began to take up a larger share of the conversation. Long-form storytelling, platform relationships, and recurring IP change expectations around scale and value emerged in the introduction of Drama Series Days (2015) and sessions such as “The Boom in High-Quality TV Series – European Drama Goes Global” (2016).

• Late 2010s - Contained experimentation. Later in the decade, experimentation becomes more contained. New ideas are tested in workshops, labs, and side programmes rather than presented as sweeping change. “Market Innovation – How to Get Things Moving?” (2017), the launch of the VR NOW Summit (2018), and “Blockchain in Motion” (2019) feel like they are part of a more cautious approach.

• From around 2018 onwards - Values moving into the main agenda. Around 2018, issues of sustainability, inclusion, and working conditions gained consistent visibility across the programme. These topics, once treated as peripheral, begin to shape how production is discussed more broadly. Sessions like “Why Diversity Matters in the Film Industry” (2018) and “Strategies for Green Production and Financing” (2018) mark this shift.

• Early 2020s - technology as infrastructure. In the early 2020s, technology is discussed less as a possibility and more as part of daily operations. Data, platforms, and automation are assumed components of the production process. “AI: Film’s New Normal?” (2020) captures this change in tone.

What EFM 2026 has in store for us

And that brings us to the present moment. There are 131 sessions taking place over just seven days.

As I’ve shown in previous research, although film industry events may at first seem to span several weeks, they are almost always focused on the first weekend.

And EFM 2026 is no exception. Anecdotally, I have heard a number of people say they’re flying in only briefly, with few expected to stay past Monday.

We can see this reflected in the event programming. Although the sister event, the Berlin Film Festival, takes place over eleven days (12th to 22nd February), the market will focus on just four days (Friday 13th to Monday 16th February).

Reading between the lines of those 131 sessions, the key themes seem to be:

1. The biggest conversation is about how to get through the next few years. The strongest theme across the programme is an industry that is focused less on growth and more on managing risk, tightening operations, and staying viable in a prolonged, tough market. Sessions: State of the Industry; State of Finance; How To Thrive in a Challenging Market.

2. AI has become everyday work. AI is something already embedded in development, financing, marketing, and decision-making workflows. Sessions: From Hype to Help: Real-World AI Use Cases Worth Your Time; Does AI Need You More Than You Need It?

3. The franchise dream is being handled more carefully. There is caution around IP expansion, with conversations focusing on when cross-media ambitions add value and when they actively weaken projects. Sessions: Hard Truths about Transmedia: What Works. What Doesn’t.; Worldbuilding Across Media

Read more

No, it’s because the movie stinks. And they know it. And they don’t want the rest of us to know it until a few minutes after the movie starts, once they’ve got our poorly spent money.

But this blog is not about guesses - it’s about seeing what the data reveals. Which means we have two questions to answer today:

1. If a movie doesn’t have advance reviews, should we use that as an indicator that it won’t be very good?

2. And if it is true, should we use this to decide what to see in cinemas?

Today’s topic was suggested by Matt Liebmann from Vista Group, who, despite being deep in planning their upcoming VistaCon 2026 conference, took a moment away to ask:

Is there a correlation between when a movie’s reviews are released and the box office?

I turned to the data to find out. I tracked the publication date and review score of 175,902 film critic reviews across 10,766 movies released since 2000.

Let’s start at the beginning and ask why distributors would want to delay bad reviews, which are going to come out anyway.

First, a little promo for a fellow film data Substacker - Daniel Parris...

Stat Significant is a free weekly newsletter with 24k subscribers featuring data-centric essays about movies, music, TV, and more.

When do we stop finding new music? Which TV shows got their finale right, and which didn’t? Which movies popularised (or tarnished) baby names? Subscribe to Daniel’s blog for free to find out!

Subscribe to Stat Significant

The advantage of delaying reviews

The theatrical market is built around a narrow window in which information and money move very quickly. Opening weekend still does a disproportionate amount of the work, especially for wide releases.

So if a large share of a film’s eventual box office is earned before audiences have had much time to talk to each other, distributors will want to protect that.

A 2006 study found that the quality of the movie had much less of an effect on opening-weekend box office than on overall box office gross. You could read this to say: the more time people have to learn about how good the movie is, the more it matters if it’s any good.

Another study in the Journal of Marketing found that reviews have their strongest impact early in a film’s run. Negative reviews, in particular, hurt box office performance more than positive reviews help it, but that asymmetry is concentrated almost entirely in the first week. As the run continues and audience word of mouth takes over, the effect of criticism weakens.

In other words, bad reviews are most dangerous precisely when studios are trying to maximise opening-weekend demand.

So let’s turn to the data and see if we can detect this effect in the wild.

Do earlier reviews correlate with better films?

Yes.

When we track the Metascore for movie reviews by the time difference between the review publication date and the movie’s release, the average score falls.

Reviews for both parts of the recent Dune duology came out well over a month before the movies were released, gaining Metascores of 74 and 79 out of 100. This built buzz and helped the movies have stellar opening weekends.

Conversely, reviews for Morbius (Metascore 35 out of 100) were embargoed right until the last possible moment, landing effectively on release. When they did appear, the response was overwhelmingly negative. The film opened on marketing and brand recognition, then collapsed once audiences caught up with the critical consensus.

Do even earlier reviews correlate with even better films?

Again, yes.

If we zoom out wider than just the week of release, we see an even stronger effect

The Metascore of movies reviewed months ahead of release clusters in the low 70s. Films reviewed in the final days cluster in the low 60s.

There are two things going on here:

Firstly, when a movie is way above average, then the senior folk at the distributor or studio will want to shout about it. In part because it’s such a rare occurrence that the sheer shock causes them to jump out of their palatial bath, run naked through the streets, shouting “Eureka! I always knew I was a genius” and also because it can form part of the marketing strategy.

Word of mouth has consistently been shown to be the most powerful force in movie marketing, so when the buzz looks good, it makes sense to lean heavily into it. The marketers will want to get as many chatty people with a platform to see the movie, in order to build anticipation. This will include “talkers“ (i.e. advance cinema screenings for the public) as well as getting professional reviewers to see it earlier.

The second big reason is film festivals. When a film premieres at film festivals, film critics get a chance to review it months before its commercial release.

In general, film festivals tend to show the better films (although not always, see point 6 here) and so it makes sense that the films reviewed early at film festivals will receive higher-than-average reviews from critics.

But even if the film is not officially “In Competition”, film festivals can be great platforms for distributors to use to gain buzz and get the word out there. Top Gun: Maverick had a glitzy red carpet premiere at the Cannes Film Festival weeks before its release, earning column inches in the celebrity-focused press and positive reviews from film critics.

So when we see reviews appearing three, six, or twelve months before release, it usually reflects a festival debut rather than a distributor taking an unusual marketing risk.

Notes

The dataset covers 175,902 individual critic reviews of 10,766 commercially released US films between 2000 and 2025, for which I found at least 10 reviews. Each review is matched to a publication date and a theatrical release date, allowing me to measure the gap between the two.

My approach weights reviews equally, without accounting for the outlets’ influence or reputation. It’s also tracking the reviews, not film averages. Both these choices will affect the numbers, but I don’t think they change the outcome.

I’m attributing almost all of the correlation here to the quality of the movie, but it’s possible that critics are punishing films (relevant reading: Do film critics punish films with bigger budgets?) because they feel snubbed from not getting a sneak peek ahead of the Great Unwashed. While it’s possible that perceptions do play a role, I very much doubt this is the core mechanism.

Wow, such a great movie. I’m never watching it again.

So much so that it has become shorthand among film fans for ‘A great movie that doesn’t attract repeat viewing’.

When you stop to think about that concept - an objectively great movie that people avoid - it’s quite kinda strange. If you like watching great movies, and you think that Requiem for a Dream is indeed a great movie, why wouldn’t you want to watch it again?

Added to that, there are loads of mediocre and bad movies that we all collectively choose to re-watch, despite knowing (either from experience or their reputation) that better movies are available.

I’m going to tackle the notion of “re-watchability” in a future article, but today I’m going to focus on the following question:

What is the element that, when present in a great movie, means that audiences never want to watch it again?

To do this, I built up a dataset of 19,821 movies which appeared at least once on 530 user-created lists on Letterboxd, centred around the theme of ‘Movies you’ll only watch once’. I then narrowed down to the most oft-cited movies and their ratings.

In doing so, I think I have worked out what it is that is driving the ‘One and Done’ approach to viewing some great movies.

The archetypal one-time gem

Let’s start by studying our principal example to ensure we know what we’re looking for.

Requiem for a Dream was directed by Darren Aronofsky and adapted from Hubert Selby Jr.’s 1978 novel of the same name. It follows four characters in Brooklyn whose lives gradually unravel through different forms of addiction.

The film is about one’s options narrowing and the inevitability of a bad outcome, even as the characters seek escape. The way it’s shot only adds to this, with lots of repetition, escalation, rapid editing, extreme close-ups, and an insistent score to reinforce the sense of momentum without release.

The movie consistently gets high ratings from people of all types.

• Critics give it an average of 71 out of 100 on Metacritic.

• 80% of the 128 critics on Rotten Tomatoes gave it a positive review.

• 43.8% of the cinephiles on Letterboxd who rated it gave it 4.5 or 5 out of five.

• 75.9% of IMDb users gave it a rating of 8 or higher out of 10.

Despite its universal acclaim, when I built a list of Letterboxd user lists along the lines of ‘great films you don’t rewatch’, Requiem for a Dream appeared on 99.2% lists.

That’s right - it seems almost impossible to make a list of one-time masterpieces and not include Requiem for a Dream. The next most frequently mentioned films were way behind, and were:

• Mysterious Skin (2004)

• Grave of the Fireflies (1988)

• Dancer in the Dark (2000)

• Come and See (1985)

• Oldboy (2003)

• Eternal Sunshine of the Spotless Mind (2004)

• Funny Games (1997)

• Irreversible (2002)

• Schindler’s List (1993)

• Beautiful Boy (2018)

• Manchester by the Sea (2016)

One theme to ruin them all

I wanted to see if I could deduce a theme running through the long list of movie titles.

My first thought was bleak intensity. Requiem for a Dream is exhausting, bleak, and emotionally punishing.

But when I looked at what made the films on the list stand out, there didn’t seem to be any difference in this regard from many other movies.

Many violent, upsetting, or emotionally heavy films attract repeat viewings, with some even becoming comfort watches for certain audiences.

• The Shining (1980) has sustained dread, violence, and psychological collapse.

• Alien (1979) is claustrophobic, violent, and relentlessly tense.

• Se7en (1995) is so bleak - to the extent I think many people forget just how grim it is until they rewatch it. (Not to mention nihilistic).

• Jaws (1975) has sustained dread, violent deaths, and existential fear.

• Rosemary’s Baby (1968) is all about intense paranoia and violation.

• The Texas Chain Saw Massacre (1974) is relentless and stressful.

So I don’t think that bleak intensity alone reliably indicates whether people come back.

And then I found a thread which ran through many of these one-and-done classics..

Powerlessness.

What Requiem for a Dream does so well is place the viewer in a position where effort and insight offer no protection. The film does not invite you to anticipate improvement, recovery, or escape. Each character moves forward with a sense of inevitability, and we as the audience are carried along.

Unlike bleak, intense films such as The Shining, Alien, Se7en, or Jaws, knowing where the story ends does not make the experience easier. With those films, as we become more familiar with them, we experience an added sense of orientation and control. Our knowledge of the movie helps us turn fear into anticipation, making it manageable.

The opposite applies to films like Irreversible, Come and See, Dancer in the Dark, and Grave of the Fireflies. In each case, knowing what is coming does not provide comfort or control. It removes any remaining distance. The outcome is fixed, the damage is unavoidable, and we are required to sit with that knowledge during the movie.

So that’s my theory - great films that people don’t want to re-watch often elicit an intense feeling of powerlessness in us.

Notes

The user-generated lists I gathered for this research all had two criteria:

1. They claimed to contain films that were good, high quality and/or admired (which I then verified via user ratings); and

2. Were films the author would not want/choose to watch again.

To give you a favour, here are a few:

• Amazing, but never again

• Best movies that I’ll never watch again

• Great films I’ll never watch again

• Movies everyone should watch once that I’m never watching again

• One-Time Masterpieces

The language people use to name and describe these lists is fascinating. They fell into three categories

1. Refusal. This was the “Never again” type, which implied that the author was adamant they were not going back.

2. Once only. This summed up the ‘single viewing’ perspective in a fairly neutral way.

3. Must see. This was the smallest group, but still present. They framed it almost as a duty to watch these movies.