Hacker News new | past | comments | ask | show | jobs | submit login
20th Century Fox Uses ML to Predict a Movie Audience (cloud.google.com)
141 points by xTWOz 28 days ago | hide | past | web | favorite | 137 comments

I know I’m not the average movie-goer, but I strongly prefer to see movies at either end of the spectrum: really bad or really good. Those are usually experiences I can fully immerse into and enjoy.

Anything in between (I.e. mediocre) does nothing for me, no matter how well produced or marketed it is or how many hundred millions it cost to make. It’s just gonna be a dud.

Most movies made to please the most people end up here: high budget, great looking, but mediocre, predictable and utterly boring.

To me, using ML to optimize movie audience targets sounds like something which will make a bad situation even worse.

I guess cinema as an art form is truly dead.

Edit: This news seems to confirm a claim made by screenwriter Max Landis[1] in a RLM podcast: Hollywood used to have a recipe for making “successful” (i.e. profitable) movies, but these days none of the recipes work anymore and they are desperate, feeling blinded, and as a result being even more risk-averse in their productions.

This definitely sounds like a desperate move.

[1] https://youtu.be/DR-Dry8Qb4A

Most movie-goers are primarily goers and only secondarily viewers. They decide to go to the movies, then they decide what they want to watch. Major blockbusters change the equation and draw in non-habitual consumers (hence their importance to Hollywood), but most movie productions aspire to little more than an appealing trailer, a bit of star power and sufficient entertainment value to avoid alienating their core customers. Retaining those customers is much easier than persuading someone like you to make a special trip to the theater for a critically acclaimed movie.

The rise of Netflix and the decline of Blockbuster has decimated the secondary market for movies, further pushing the industry towards conservatism. It's much harder to convince people to pay to see a quirky indie picture or a sleeper hit at home, so movies that don't do well in the opening weekend are pretty much a dead loss. As with Spotify and the music industry, $12/mo subscriptions can't make up for the loss of $9 cinema tickets, $4 rentals and $15 DVD sales.

Cinema isn't in a healthy place right now, much to the benefit of television - the economics for broadcast and streaming TV are much more favourable and there's an influx of talent.

> Most movie-goers are primarily goers and only secondarily viewers. They decide to go to the movies, then they decide what they want to watch.

Any cites for this? I am very curious to read more about this.

Me too - it's a really insightful statement. I'd never thought about it before, but in social settings, the actual film watched is frequently an afterthought; it is secondary to the plan of 'going to the cinema'. Plans changing because someone is late is another factor. I really enjoy when this website makes me think of something obvious I'd never considered!

I was thinking about it from the other way around: because I'm not sure I know anybody who would go to the theater without really wanting to see the film being shown.

I don’t think that’s exactly what is being talked about; rather, people say “let’s go see a movie”, and then look at the movies app to see what’s in theatres lately. Of course they won’t end up going if there’s nothing worth seeing—but their criteria for “something worth seeing” is a much lower bar than it would be if they didn’t already want to “go see a movie.”

When a new movie comes out, I say either “that looks like it would be a good movie to watch in theatre”, or (far more often) “I think I’ll wait for that to come out on iTunes/Netflix/etc.” When I’m making that decision, the bar is high. But that’s because watching-at-home really is good enough for the vast majority of movies—it doesn’t remove anything from the experience.

I understood what was being said. I am questioning the point of it. Everybody I know who I've ever talked about going to a film with (and my girlfriend really likes going to the movies so it comes up with her and her friend group a lot) always always always has a movie in mind when they come up with it. The most recent example for them (I didn't go) was the new Spider-Verse movie. They weren't going to see "a movie", they were going to see "that movie", and it was from that movie's release that the impetus originated.

I'd like to see evidence of assertions to the contrary, because to me they don't jibe with reality.

Dates. The time when people go to the movies to “go to the movies” is when they’re on a date. (Not just at first, but also later in the context of a steady relationship where they want to see a movie together “because we haven’t seen a movie together in a while.”)

For me, seeing a movie in a theatre vs. watching at home is a bit like going to a restaurant vs. eating at home. I used to “go to restaurants” to try specific things I’d heard about, when I was single. Now, though, I mostly only go to a restaurant to go to a restaurant, because that’s a nice thing I can do with my wife. With restaurants, this is the difference between choosing by word-of-mouth (“they have great X!”) vs. choosing by reviews (“this restaurant seems like they’d have something we’d enjoy, let’s go and find out.”) Movie theatres don’t really have the same element of diversity (unless you’re going to independent cinemas), so it’s more just a choice of whether to go to “the movies” at all.

On the other hand, everything I said about restaurants applies exactly to the other kind of theatre. There’s a Shakespeare troupe here (Vancouver) that, each year, puts on three of the bard’s plays (several showings apiece.) You can certainly want to “see The Tempest”, and therefore go watch them perform it; but what I hear much more often is people wanting to “go to Bard on the Beach” (the troupe), and so then they look at what they’re putting on this year to see if it’s anything they could stand to watch. In this context, the theatre troupe is exactly like a restaurant: they want to go to it, as long as it has something they’d be okay trying.

Agreed I don’t know people who plan to go to the theater as a general destination rather than to see a particular movie. But I’m not an avid cinema visitor.

I think you’d have to be comfortable seeing big budget mediocre films if that’s one of your hobbies. There's a ton of bad movies out at any one time. Or rather a ton of mediocre (at best) ones.

This could be a HN demographic thing. The comment definitely piqued my curiosity, too.

Interested in others thoughts. I think the sibling commenter may have got it right with the fact people make a conscious decision to go see a movie beforehand as an outing, rather than turn up at the cinema and see what's on.

>I guess cinema as an art form is truly dead.

You say this because of something 20th C Fox is doing?!

May I suggest not watching any movies from the US for a year or 2, at least. There's a whole world out there. e.g. France, Spain, Argentina, Germany, Italy, Japan, Russia, Korea, Brazil, Iran, India, Denmark, Norway, Mexico...

(A page on my website listing outstanding movies/series I've seen recently from each of those countries and others, I don't know a better place to point you to jump straight in: http://www.adamponting.com/movies/ )

It's... An amazing list, thanks for sharing.

Can I ask two things:

- Where do you get your movies?

- Do you watch a new one daily to have seen so many?

So glad someone got something from it!

Hehe yeah, does look like a lot. But 2012-now is..~2800 days, probably there's less than 1000 movies/episodes on there, I haven't counted. Recently watching a lot of series...I like the longer form. I moved into a house in 2012 where they introduced me to torrents - before then I'd hardly seen any great movies, so that was really amazing watching all the great movies and directors I'd always heard of but never checked out. (Had only downloaded books and music before that I guess!) Used to often watch 2 or 3 movies on weekend days. Lately a couple of episodes of a series (or more if bingeing) each night, or a movie. I did put a lot of hours into researching, reading reviews, finding good things to watch, so that we didn't see many that weren't good.

Finding obscure movies online is sometimes tricky (i.e. can take hours), also tracking down good subtitles can be hard, and often have to resync/shift them and/or fix them up. Being able to adjust audio and subtitles while watching with VLC is a must. I get them from:

1. torrents. Usually only works for popular/recent/famous stuff. Sno's torrents with hardcoded English subs are great, have watched a lot of those - mostly series, some movies.

2. movie and/or series sites like primewire, the now-defunct alluc, (which was amazing, finding obscure tv series episodes is much harder now), zocine.com, megashare.bz, www1.swatchseries.to that have links to online files. These come and go over the years. Adding "watch online" to google searches finds sites that the links to movie files can extracted from, either manually from the HTML or using TubeOffline-type downloader sites.

3. video sites. Google the movie/series name - often movies/episodes are on youtube/vimeo/dailymotion, especially when I forget to look for them there! Archive.org has a lot of stuff. (e.g. recently watched White Zombie (1932), the first zombie movie, and The Best Years of our Lives from there)

Not uncommonly an obscure movie will seemingly only be online in one hard-to-find place. (e.g. The only place I could find one classic Australian movie was on someone's page on vk, russian facebook!) I have a startpage with searches each of those sites one click away, so I only have to type the name (& mostly year) once. Whenever I get something from a site I add the site to my start page to easily search it in future. Also my computer is a Mac from 2005 which doesn't make it any easier! Good luck.

Thanks for the detailed answer!

I'll dive into your list over the weekend, and if I find some missing movies I enjoyed, I'll send you an email with a few suggestions for you to watch :)

Ah thanks!, just saw this now.. Also I have a link on that page to IMDb lists I've found useful, they're a great place to find new ones. Or on the IMDb page of a movie you love, look at the lists with that movie in it.. Also, I didn't mention it, but the quality test is to read a couple of pages of user reviews on IMDb. Usually get a good idea whether I'll love it from that. (And for a movie to have 50% 1 star and 50% 9 star reviews is a lot better sign than all 7 stars.)

> I guess cinema as an art form is truly dead.

I think cinema as an artform is thriving, just outside of the mainstream Hollywood, etc studios.

I agree. Fox is a factory that happens to make movies. A corollary here would be to say that "fashion as an art form is truly dead" because Wal-Mart sells mass produced graphic tees for children.

I have your same criteria. I justify it this way, "We live in an era where the amount of 'pretty good' media being produced outstrips my ability to consume it -- even if I were to dramatically increase my viewing/reading/playing time. Therefore, I try to select only the media that's truly exceptional. Exceptionally good or exceptionally bad."

You might be falling for the branding friend!

Studios have had a major vested interest in proclaiming their product is “Glamorous,” and “Unachievable in lessee glory.” Which in turn justifies the high budgets and higher ticket fares.

But they are a pipeline, that pump 10-300 million dollars into a project that they won’t see a return on for 3 years at the shortest; starting from the time you pay a scriptwriter following through the Opening weekend.

It’s in their best interest to tell people it is Once in a lifetime experience and then put out something that is going to me market accepted.

On the occasional times, it is TRULY once in a life time (Mad Max: Fury Road for me), it hits us in the soul, and draws us back in.

Worth it, to me.

Once people realize the stupidity of hoarding and analyzing all the data that we do, I think they'll back off a bit. Too often these processes yield suboptimal results. We're in a dark age now because data science is currently quite sexy, and intuition decidedly not

Still, we’ve got to go through the medieval era to get to the enlightenment and modern era.

Hopefully these big companies wasting millions on this mostly mediocre returns will find some gold for the future. Eventually it will be very useful in a more general sense, but it’s going to take time to get there.

All we can hope for now are mainly niche wins for an ever expanding set of problems. Who knows if big budget films will be one. I’m not yet convinced by this article but at least they are trying and have the R&D budgets to fail a few times.medieval

I cannot remember who said this, but I mostly agree with the sentiment of "there are no good or bad movies, only boring or interesting movies". I watch a ton of low-budget crap-films, because I find them funny, and I watch a lot of movies that are really interesting (either directly or intellectually). A movie that is boring is the worst kind of movie to me.

A boring movie is something that I genuinely felt wasted my time. I got nothing out of it, and I couldn't even get a lot of enjoyment out of making fun of it. The perfect example of a movie in this category is the 2009 movie "9" (sorry if that's your favorite movie).

> "there are no good or bad movies, only boring or interesting movies"

I appreciate the sentiment but in this case, wouldn't that make boring movies bad and interesting movies good? (Assuming, of course, you don't want to be bored...)

Not really. Jurassic Park is a good movie, the framing, editing, script, fx work, performances, music all come together really well.

Birdemic is a bad movie, in that none of the individual components are good and they don't complement each other well at all.

I'd argue that both are interesting

So there are good and bad movies, and orthogonally, boring and interesting movies. :)

Yeah, maybe I should clarify; there are "good" and "bad" movies from a theoretical level. My enjoyment largely rests on how interesting I find them. Objectively terrible movies like "The Happening" or "The Room" or "Samurai Cop" or "Troll 2" or "Angry Red Planet" are "bad" because whomever made them seems to have no understanding of film-theory, or what it would take to make a "good" movie, but I find them endlessly entertaining...like watching a trainwreck

The irony of it is that if they had tried to make those moves "bad" on purpose, they wouldn't be interesting at all (at least not to me), and they would just be "bad" on every level for me.

In fact, even when there is sincerity, it's not a truly deterministic system. I think that the first Troll movie is technically objectively worse than Troll 2, but I find Troll 1 to be pretty boring, in addition to being incompetent, while Troll 2 is silly and an utter blast to watch.

I have the same taste in movies as you (awful B-grade or cinematic masterpiece, nothing in between), but I have the exact opposite conclusion regarding movies as an art form being "dead".

Just like with music, the technical barriers to entry to make a film or a song are much lower than before, and the market is very saturated as a result. The issue is no longer "is there enough content that I would enjoy" but rather "how the heck to do I make a choice what to watch?".

Agreed, every movie I've been to for the last 2-3 years have been boring not terrible, just utterly boring and a waste of time and money.

Compared to other mainstream art forms, movies have a higher barrier of entry and that translates in to smaller influx of new and interesting concepts.

Then you are going to the wrong movies friend.

Which is to say, you are victim to assuming the mega complex Movie Theater, powered by the mega complex of Studios pumping out advertisements are the only option available to you.

If cinema is a desirable experience to you, seek out film colleges, independent theaters, movie rental spots, and YouTube/Vimeo/reddit. You’ll find a bunch of crap, for sure, but it will be new and interesting crap. You’ll explore your own mind trying to define why you don’t like it. Eventually you’ll see something truly moving and that will be amazing!

Or, If you justly say “I don’t have time or energy for something I’m not that passionate about,” continue to go to the AMCs. Just recognize, much like McDonald’s they (and the studios) have massive overhead and massive interest in not losing money. No one takes the person complaining that McD’s isn’t gourmet very seriously.

(The most moving zombie movie I ever saw was 15mins on YouTube and had 3 zombies in it..)

Movies used to be better without having to go for the art school/criterion outlets. Cf the entire pre-Star Wars 1970s oeuvre which mostly managed to be relatively highbrow and entertaining in a way that they haven't since then.

That's because the duds are filtered through time and forgetfulness. Rose tinted glasses of memory.

Nobody remembers the boring movies from their childhood.

Erm I didn't watch those movies when I was a kid, and I certainly wouldn't be able to appreciate them if I had. I watched almost all of them as an adult within the last 4-6 years.

Me too, but that's not the point. We are not watching a random sample of old films, we only watch the old hits.

We are watching a random sample of present films.

I don't think I am watching the old hits! I watch a lot of forgotten trash, and I like that stuff too. It's still way better than capeshit or whatever it being hawked these days. Video nasties by Hammer, weird pirate movies written by Peter Benchley, Charlie Bronson movies, Oliver Reed movies, Doug McClure movies...

A lot of the older film noir stuff I watch is definitely filtered by time: but you can pick any random garbage from between 65 and 75 and it's likely to be decent.

This! Not to mention that at the time, people HATED Ewoks.

"Success in the movie industry relies on a studio’s ability to attract moviegoers—but that’s sometimes easier said than done"

No it isn't. Just make a good movie. Forrest Gump didn't need any machine learning.

The only reason someone might use this is when your strategy is "spray and pray", create crappy movies that insult everyone's intelligence and you need to figure out is it worth launching at all.

> Forrest Gump didn't need any machine learning.

No, it just needed a massive demographic the studios could exploit the Hell out of by taking a semi-popular novel and Spielberging all over it.

You have a good point, but you used an imperfect film to demonstrate it. Forrest Gump was just about as cynical a film as you could imagine, and succeeded precisely as predicted. That kind of cynical filmmaking, explicitly targeted at demographics with earnings projections in place before casting begins, is precisely what this AI-driven concept will end up with at its worst.

A better example would be the works of the Coen brothers, or David Lynch: Even if you don't like them personally the fact remains they have an audience, and they most certainly weren't made to order based on market forecasts. They represent the auteur theory in action, being the products of distinctive creative minds who impressed their equally distinctive visions upon them.

This is being a bit too "edgy" for me. Not to defend Forrest Gump as an all-time great or anything, but is there really something wrong or unsavory about its conception?

What you call "exploiting a demographic" is more like "having a target audience instead of trying to be all things to all people". Sure, Boomers are a huge generation. But they were entering their 50's when Forrest Gump came out... and studios usually aim at 18-34 year olds because that's who spends the most time and disposable income at the movies.

What you call "Spielberging" is probably just a combination of cinematography, soundtrack, and accessibility over hipster snob appeal. In other words, good mass-market filmmaking.

Maybe this particular movie isn't your cup of tea, but there's nothing negative or cynical about this. "Cynical" would be hijacking HN threads to bring up urban walkability, UBI, random potshots at Baby Boomers, or any of the other standard hot buttons that typically guarantee upvotes and validation.

I believe the cynicism would be when a studio chooses demographic and marketability over art.

There is no "best" art (sorry, high art world), but there are absolutely choices you can make for marketing rather than artistic reasons.

I like Robert Zemeckis and Ron Howard films, but they're boring ideas, well-executed. And I have a sneaking suspicion that given their copious directorial output and uniformity of end product... some of those choices were made for monetary rather than creative reasons.

Nobody says "I want to make a mind-bending Dune film, get me Robert Zemeckis!"

> "there are absolutely choices you can make for marketing rather than artistic reasons"

That is fair. I believe it's also the case that when evaluating art, it's easy to conflate those reasons with one's own POV and demographic biases.

In other words: Forrest Gump is cynical pandering to Baby Boomers... but [this-year's-Marvel-superhero-movie] is an artistic gem, and The Dark Knight or Black Panther got robbed at the Oscars, etc.

That's an odd way to characterize it, IMO. Like if someone wanted to argue that Zemeckis makes substandard movies because he wants to profit from it, I should be able to argue that the Leanardo DaVinci painted substandard art because he profited from it. 100 years from now, nobody will care if Zemeckis made money, because he'll be dead. The legacy is what remains. The art is what is left behind, and what is left behind is better preserved if more people like it. I would argue that the value of art is not a pure measure of quality, but more like the product of quality and outreach.

I'd say it is unsavory in the same way today's movies are "manufactured" to extract the most value from american and chinese markets. Where decisions of film locations, dialogue, plot, etc are driven by marketing and political aspects of these two major markets. And the "art" comes after the market research.

"ars gratia artis" doesn't apply to Forrest Gump. It was manufactured out of marketing data. It was the 90s version of most recent nostalgia film "Ready Player One". I know it won a bunch of awards ( which I put no credence or value on ) but it is film saccharine. Forrest Gump is one of the forefathers of the current movie industry.

Just because a movie is created because of cynicism on the part of the production company does not mean that it is created because of cynicism on the part of the filmmaker, crew, or editors.

A filmmaker will flagrantly appeal to the cynicism of the producers in order to produce a film they believe in wholeheartedly. Forrest Gump is probably a great example of it.

Spielberg and Ron Howard definitely make middle-of-the-road appeal-to-everyone films.

But so did Shakespeare. Middlebrow and excellent are not mutually exclusive terms.

Oh, and Michael Bay is represented on the Wikipedia List of Film Auteurs page.[1]

[1]: https://en.wikipedia.org/wiki/List_of_film_auteurs

>massive demographic the studios could exploit the Hell out of

who? readers of a novel?

Baby Boomers! The film Forrest Gump was basically Baby Boomers: The Movie.

It was a Greatest Hits album of everything the stereotypical Boomer would consider great and memorable about their childhood and young adulthood.

My wife and I are 47 and love it. Our teenage kids enjoyed it when we watched it with them recently. My dad, a boomer, actually doesn’t care all that much for it. It happens to be a good story irrespective of when you were a child or young adult. Shrug.

Certainly doesn't hurt that Tom Hanks attracts all kinds

I'm not sure what you're trying to say here - there's nothing wrong with enjoying it, there's nothing wierd about younger people enjoying it and that still doesn't really change the fact that movie targets nostalgia of a certain demografic. Considering it's success, it does that very well.

I saw it in my teens and loved it. It was a film that could appeal to all.

Disturbing modern idea that everything has to appease only a certain group of people, others need not apply.

What in the world are you talking about?

Gump was Zemeckis!

those 2 have quite a similar approach to movies AFAIK

Spielburg produced a lot of his films, so that would make sense.

As with most simple recipies "just make a good movie" is unfortunately neither sufficient nor always the best path. Word of mouth on movies is far more likely to have a negative impact on theater going decisions than positive impact (if your friend says it sucks, you are far more likely to switch from go to no go than you would be to switch from no go to go if the same friend said it was awesome). In practice, what this means is that because positive word of mouth is so minimally impactful studios need to focus on high marketing value films rather than high story value films, because high marketing value films cause potential audience members to see a trailer or an ad and say "I'll totally go see that!" Yes, negative word of mouth can and will reduce the number who actually go see it, but final viewership is higher if you start with a huge population saying "I'll totally see that" than if you start with a "meh" response to the initial ads and try to grow viewership through word of mouth about how awesome your film is. Given these observations, Hollywood has made the rationally and financially correct but unfortunate for audiences decision to focus on marketability over enjoyability. These economics are quite different for entertainment as a service providers like HBO and Netflix, with the obvious results in quality of output.

The alternative is to build up a reputation, so when a new $studio film comes out, people are excited about it because $studio has a history of making good films. I think Pixar is/was a good example of this.

Every studio that has focus on children is smart marketing.

Children > parents.

If you are a parent, don't you go to a child friendly restaurant that is reasonable good instead of one that is very good but not child-friendly.

That only works if the films are still good though. Pixar has had a couple of bombs, The Good Dinosaur for one.

No it isn't. Just make a good movie.

Lots of "good movies" (using any definition you want) fail to find an audience. And even more mediocre movies make huge buckets of money. If your goal is to make money then just blindly making "good movies" is probably not your best strategy.

> If your goal is to make money

That's one of the reasons why we have so many poor movies (the other reason is that making movies is hard). If your only consideration is making money, you'll produce whatever crap is cheapest to make that still makes reliable income. This is why e.g. consumer market is full of barely-working near-disposable products which existence is a pure waste of natural resources.

Seriously. There's no rule that says that the most profitable product is also the best for the user. In fact, it's almost always the opposite, and A/B-testing things to death leads to the former type of product.

This is pretty much entirely wrong.

Studios want what's "proven". It's why they will gladly feed in ML data to get shitty focus group type movies. They will then pour huge amounts of money in to try and get back even more.

Compare and contrast "good artists" where most really great artists died as paupers and found recognition long after their deaths.

Depends on your horizon.

> Depends on your horizon

Box office returns dominate the return for a film, set the license price for long term post-theatrical deals, and decay with a very fast exponential from the opening weekend take. It is extremely rare in modern times for a film to turn a poor opening weekend into a high long term return. The film market simply doesn't work that way however nice it would be if it did.

I read the comment you're responding to a bit differently.

I think most people can enjoy a really bad movie. Some t-and-a, explosions, a bit of CG - it's all good fun even all you can think about after the movie is how awful it was. But when really bad movies are all there is, they quickly lose their charm. Instead of being one off guilty pleasures, they become the movie industry. And that's about the point you stop watching movies at the theater.

I think this describes many people's transition away from movies. The point of this is that pumping out crap may be optimal in the short run. But in the longrun you end up with suboptimal returns, even when profit is your one and only motivation. Incidentally the same story is true and also being played out in various other industries.

Optimizing for next year's quarterlies is often not the same path to optimizing for earning the most money in the next decade.

> No it isn't. Just make a good movie

"Good" is not a single-variable problem, you know? Sometimes you have a good script, but you don't have the perfect cast. Sometimes you have decent actors and all but almost no marketing budget so nobody hears about your movie. There's so many things that can go wrong even with the best intentions when trying to make a good movie that "just make a good movie" sounds really naive.

If it were that simple, every "good" director would make great movies one after the other. I struggle to find more than a few who could pull that off (and even then, that's highly debatable), among the the tens of thousands of directors who have been active in the past 50 years.

> Just make a good movie. That seems obvious- but good movie does not mean high grossing blockbuster. Shawshank Redemption made less than the Bee Movie.

Shawshank Redemption joins Office Space, Fight Club, and It's A Wonderful Life in the ranks of films saved by being shown on TV (whether via cable, DVD, or obsessive rebroadcasts over the holiday season before It's A Wonderful Life got snatched out of the public domain) after having bombed in theaters. They eventually found an audience even if the vagaries of the theater world kept them from making much in box office sensu stricto.

There are movies I want to see multiple times and movies I truly loved but only needed to see once. I would assume block blusters fit the "multiple times" category which is why they make so much money, because people go more than once.

A drama like Shawshank Redemption (a favorite of mine) is not the type of movie I would have paid to go see multiple times in the theatre. Where as I saw The Matrix 6 times in the theatre back when it came out. Every friend who wanted to go see it I went with.

Maybe they needed to get a bigger marketing budget. Now, if only the majors could predict which movie need to be marketed how...

Not all good movies can have good/effective trailers made for them, and not all word of mouth can overcome a bad trailer.

Forrest Gump had a $55 million production budget and a $71 million marketing/advertising budget:


That's unfortunately totally wrong. Good movies rarely get released, for decades.

I'm professional film critic and I have a good idea of the data, because I analyze the feedback from festivals and industry experts, who decide what gets released and what not. From the Top 10 typically only 3 get selected.

The selection criteria has nothing to do with the quality of the movie and rarely the quality of the trailer, even if it's more important.

Most important are the names. Hollywood divides the Actors and Directors into A+, A, B+, B and C. Without an A name you'll have a hard time to get financing and proper distribution.

2nd most important criteria is the marketing power. How much money will be spent on marketing. If the distributor is not sure he'll try smaller tactics, like smaller releases, with the risk to earn much less. Most important is only the first weekend. Really good movies without names rarely exist, which can drive marketing by its own. Only these movies can survive a bad opening and raise slowly over the weeks to get profitable.

Hollywood knows exactly the triggers beforehand, that's why they can get away with so many formulaic and simple movies still bring profitable. There's not much need for ML, just counting the numbers and trusting the experts is enough. In this case, if a trailer works for which group is certainly helpful though. With trailers you get better feedback than from test screenings, but not much better feedback than from experts.

I think the examples to watch are the many good movies that were not profitable and a decade later are "cult movies".

Tell that to Dredd.

Solid advice. On a similar note, if you want good software, you don't need fancy unit tests or type systems. Just don't write bugs.

No, the correct analogy would be don't create crappy products and use ML department to predict how many people would buy it anyway.

> don't create crappy products

Exactly, just don't write bugs.

The same way Michelangelo sculpted David, he just removed the marble not part of the statue.

Six terminators, Seven transformers, eleven Madea films, and nine aliens films. If the strategy is to waterboard us with progressively more pointless sequels, no ML in the world is going to improve the outcome when the writers have clearly checked out.

Here [1] are the box office results from 2018. Looking over the top 20 highest grossing movies I count 13 sequels/franchise movies, 3 standalone adaptations of existing IP, 2 remakes, 1 movie based on a well known true story, and only a single completely original movie. A company like 20th Century Fox is trying to make money and not necessarily make art. It is clear the best way to make money is through sequels and existing IP. Blame your fellow moviegoer for the repetition.

[1] - https://www.boxofficemojo.com/yearly/chart/?yr=2018&p=.htm

A Quiet Place outsold Mary Poppins Returns? This can't be right!

At least these are "sequels". Reboots are even worse in my book.

I think the issue is when remakes happen in quick succession. When generations go by magic like The Fly, Oceans 11, and the greatest remake of all time, The Thing, happens.

I... don’t think any of these movies have remakes?

If by the thing you mean the 2011 film, that was a prequel. It had few redeeming qualities, but thing it did unreasonably well was faithfully reproduce specific elements that were shown to have occurred previously in the first film.

The 1960 film Ocean's 11 does have a remake.

Those are all remakes. 1982 "The Thing" is a remake of the 1952 "The Thing from Another World".

It could be argued that the latest Terminator movie was an attempted reboot, since they created several new timelines, of which many completely invalidated the timeline from the original movies.

(And iirc it tanked.)

Don't forget Terminator: The Sarah Connor Chronicles, which already established multiple overlapping timelines, with people in the present from different futures.

Reboots can work well. Batman Begins, Casino Royale and Dredd come to mind.

They'll make money anyway. People will pay to watch those, so they have no incentive to do otherwise.

Like with every product that people refuse to go without no matter the quality.

Yeah, it feels like a little kick of novelty is way better than not going to the movies at all.

Why the need for a writer? The ML should be able to produce scripts of equal quality.

Why the need for actors? The ML could be trained on millions of hours of existing movies and produce fake, convincing personas. Why the need for spectators? The ML could analyse the movies and produce fake, convincing reviews.

I would love to feed GPT2 with movie scripts when it's completely open source.

Tired that (machine learning) suggestions in Amazon, Youtube, Instagram and other sites mean showing me more of the same. Is this the fabulous AI revolution?

Yes, I like photos of food, but I can also enjoy other types of photos and until I don't start searching for them, Instagram (or whatever service) doesn't list them. What a great AI!

This is like that article that talked about increasing extremist political videos being recommended depending on your views habits (Youtube).

I understand, no data can create predictions. But tastes are so depending of the context you grow up in that it is normal not to be able to predict that. And that uniqueness is what makes a person great.

I guess 20th Century Fox Uses Statistics to Predict a Movie Audience doesn't have the same ring to it.

But I bet the 21st Century Fox could do it better.

Lol. Well, you see, there are so many complex non-linearities here that it wasn't possible until they could use a 20 layer deep neural net /s

> The top audiences for Logan were actually a combination of superhero (which we already knew) and “rugged male action lead” (which we didn’t know with certainty). This can be better seen by noting that key “rugged male action lead” predictions like The Magnificent Seven (in blue above), John Wick (in green above) and Terminator Genisys (in blue above) were also present in the top 20 list of actual audiences. This result is a win-win because the new audience “adds” to the core superhero audience, and can potentially be used to extend the reach of the movie beyond that core audience.

I wish they'd include saturation and release dates as factors as I'm sure releasing a new rugged make lead superhero movie every week would have diminishing returns at some point.

> I wish they'd include saturation and release dates as factors as I'm sure releasing a new rugged make lead superhero movie every week would have diminishing returns at some point.

Hollywood has seemed... constrained from noticing this.

Nope, it just financially works so why do otherwise. People still pay no mater how many bad movies they got fed with.

I guess the ML doesn't read the news, as there was a recent release stating that movies with strong female leads make more money [0]. Seems like the ML would figure that out, and lean heavily against 'rugged male action lead' dominate movies.


Those diminishing returns haven’t seemed to happen yet. What is it this year... Avengers vs. Batman?

Those aren’t “rugged” male leads

One of the reasons why there are so many sequels and less unoriginal movies from big studios.

Big franchise names almost always perform according to expectations, even if the movie is bad.

It's actually fairly rare for an animated sequel to outperform the original.


Incredibles 2, Shrek 2, Spongebob 2, and Despicable Me 2 were really the only ones worth making, vs gambling on a new property.

A moderately successful sequel to a highly successful original could underperform that original, but still outperform the average original.

If "Kung Fu Panda 1" makes $273 million and "Kung Fu Panda 2" makes $187 million that's lower - but higher than "Pokemon: The First Movie" which only made $153 million.

Maybe domestically? Toy Story 2, Monsters University, and I'm sure others outperformed the original when you compare their global box office numbers. https://www.boxofficemojo.com/movies/?id=monstersinc2.htm (nb foreign BO is 64% of the haul).

Personally I just consider every Pixar movie a sequel to the previous Pixar movie. I treat Pixar more like a director or author. The stuff may be different but I'm going to buy the next Neil Stephenson novel even if it's a cook book.

That was kind of my point. Pixar or Dreamworks or Disney Animation or Illumination as a brand are bigger draws than being a sequel. "I want to go see the new Pixar movie."

This is a good point. There is some sentiment in this thread, and among the public in general, that Hollywood has lost its touch and is unoriginal with the deluge of remakes and superhero movies. I don’t work in the industry, but I think Hollywood knows precisely what they’re doing.

A primary concern is how you differentiate your product in the age of Netflix. If you want to make an original and complex character driven work, you’re better suited with a Netflix series. Movie studios therefore gravitate to producing works uniquely suited to highlight the things that can’t be done on the small screen and simply must be experienced at a theater. Thus, they continue to churn out special-effects laden blockbusters based on familiar IP because it makes sense for their business.

There was a great AI Today Podcast interview with 20th Century Fox CTO Hanno Basse on how AI is being used quite intensely in the entertainment industry: https://www.cognilytica.com/2018/11/07/ai-today-podcast-62-a...

A data science consultancy I used to work at did box office predictions for a major movie studio, and the fanciest techniques you could dream up gave only a slight lift over the https://www.hsx.com/ prediction market, which provided the overwhelming majority of the model's predictive power.

> When combined with historical customer data, sequencing analysis can be used to create predictions of customer behavior.

When not combined with sequencing analysis, historical customer data can be used to create predictions of customer behavior.

It watches movies and does classification which all studios do anyway for decades on everything they and their competitors produce. I'm not really seeing the value add here apart from having a "data science" team to excuse your flops.

Here's an orthogonal idea which would make a great project - scrape IMDB for all creative roles in the credits - actors, directors, writers, set designers, make up, music, etc. -- all of them. Regress them against budget, revenue, margin and revenue/budget.

Importantly it would allow for groundbreaking and genre bending movies to be assessed.

See who's adding value and who's not. And see how well that predicts your outcome.

With enough orthogonal ideas you'll have a well rounded approach. Why would 20th Century Fox advertise their best idea?

JJ Espinoza goes into detail about the system on the twimlai podcast - https://twimlai.com/twiml-talk-220-building-a-recommender-sy...

Netflix is doing this at scale for all the content that it is on its platform, where the audience is the number of Netflix member that have watched (or will watch) a tv show, movie etc. There are limits of predictions that are reached quite fast. For example, how much money you spend on the production for an original show is the best predictor of the audience (more money spent means more promising plot and actors and directors, more marketing and so on). All the fancy AI, Deep whatever are largely irrelevant and a waste of time and money, but can be sold very well internally and externally, at least for a few more years.

This strikes me as being classically anti-inductive: if they figure out the "formula" for a successful movie, the formula will become useless as soon as they try to apply it.

And this is going to ruin the current movie industry if it gets any traction. Now that technology is highly available to everyone, this is going to cause two major streams of movie making - industrial/predicted and independent/auteur. Those exist already of course, but indie is a niche. Guess who is going to remain in business if the movie industry collapses?

The indie lives if the industry dies. The industry will not die but it may evolve.

Netflix has been doing this for years. It's one of the inputs into how much to pay to license a film or TV shows (but to be clear not the only input, at the end of the day it's the buyer's gut instincts that determine what to buy at what price, but the AI is a tool to help them make that decision).

For those interested in the economic factors driving media production and distribution, Matthew Ball [0] is a great analyst to follow.

[0] https://redef.com/author/56d1e0f14f16158e36008235

This is the reason 90% of their budget goes to comic book franchise movies. I guess it also helps that the simple-formulaic plots translate easy and rarely delve into any social issues that foreign (e.g., PRC) censors would find objectionable.

I bet this is one example of many on how they use ML. This is probably their least capable predictor- but a good example to signal that 20th Century Fox has entered the 21st Century.

Maybe someone here remembers the name but I read a short SciFi story about someone who edited a mediocre movie into an amazing movie, posted it online, got tons of hits, was offered a job by the movie company. When she arrived it turned out they wanted her to be part of an audience of people with good taste. Some AI made the movies, they'd play them to the audience with tracking equipment on their heads. The AI would use the info to adjust the movie. Their memories of having watched the movie would then be erased. Repeat several thousand times until all readings are high. Release movie.

Why wait for a movie to be produced before running ML on it to predict audience?

Why not just run it on SCRIPTS to see which to produce?


Wow I didn't know you needed ML to decide to copy what disney and marvel are doing

Using ML to speed up the process of making bland blockbusters!

I'm a bit disappointed they didn't use some old movie terms ... maybe it's in the code such as a "LEGS quotient". They used to ask if a movie "had legs" because legs puts butts in seats.

So now we get even more of the same but machine certified?

if you apply machine learning to a creative effort you will only continue to imitate what has already been done

google built the awesom-o 4000?

we're going to be getting a lot more adam sandler movies

is this why everyone makes shitty movies these days?

Now every 20th Century Fox movie will be full of violence and sex.

Ops, It looks like the ML is already in production!

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact