'Why the algorithm that promised to save Hollywood destroyed Relativity Media'

[Terry VerHaar]

Elsie - a couple of notes related to your comments...

The data this modeling looks at is actual behavior - not polls. It looks, in the aggregate, at what people actually did; i.e. what movies they saw, not what they said they would do.

And - I am not sure how the success of a movie would be "an individual outcome" since it is based on the the behavior of millions of people.

The type of "unusual" occurrence at the last hour would be considered the "black swan" I discussed above and would not be predicted by this model (nor by any decision making mode).

Finally, yes, if a "formula" for successful movies were to emerge as a result of using this type of tool industry-wide, behavior would probably shift away from that particular type of film. The good news is that the analysis would begin to recognize this as it happens. The tool itself is, by it's nature, dynamic.

And - "hints" of success are exactly what is being looked for here. No tool that assists in making decisions about the future can offer anything more.

 

[Elsie N]

Elsie - a couple of notes related to your comments...

The data this modeling looks at is actual behavior - not polls. It looks, in the aggregate, at what people actually did; i.e. what movies they saw, not what they said they would do.

And - I am not sure how the success of a movie would be "an individual outcome" since it is based on the the behavior of millions of people.

The type of "unusual" occurrence at the last hour would be considered the "black swan" I discussed above and would not be predicted by this model (nor by any decision making mode).

Finally, yes, if a "formula" for successful movies were to emerge as a result of using this type of tool industry-wide, behavior would probably shift away from that particular type of film. The good news is that the analysis would begin to recognize this as it happens. The tool itself is, by it's nature, dynamic.

And - "hints" of success are exactly what is being looked for here. No tool that assists in making decisions about the future can offer anything more.

Terry, my poll reference was meant not to show that data isn't viable, but incomplete.

As you say they track how the public has been acting, but I'm saying to really get a good feel for an individual film's potential they have to track a director's performance, the scriptwriter's performance, the actors' ability to perform in their role with this group of cast members, etc. etc.

Now these things actually are addressed for the most part by humans who are putting the movie together. But that puts probably the most important part of the equation on the ability of the humans involved to get all these things right.

 

[Hrvoje Simic]

I would hardly call analysis of large amounts of dynamically changing information on an ongoing basis, "fixed."

Obsessing over "analysis of large amount of dynamically changing information" in this context would be a compensation for the lack of inspiration, vision, experience, teamwork and basically everything it takes to do something creative and highly complex properly, to leave behind something of value.

All of which is a much greater challenge than digging through obvious, instantly perceivable and quantifiable realm and speculating within safety and comfort of an armchair.

Forest for the trees.

Human brain is already "analyzing large amounts of dynamically changing information" every day, and contrary to notions promoted by idiotized and further idiotizing status quo, does it far better than any piece of electronics.

When used adequately.

Nor is anything "fixed" when the assumptions you apply and the data they uncover are constantly in flux.

Fixed is quantitative basis used for making predictions (speculations).

Gibberish from a deleted line in a future Star Wars movie? LOL

No.

Principles presented without the unnecessary waste of words on something pretty clear to many, translated into a particular language, writings relying on the premise that the reader, besides the ability to focus on the topic, possesses the capacity and will for achieving understanding, without a fantasy land designed interpretative prism getting in the way.

When it comes to the supposed inability to use past behavior to predict future occurrences, I'd ask - what other means do you employ?

You are asking a question on a basis made by generalization and you fail to see what can be predicted and what cannot.

One cannot predict director's script execution, producer's project decisions (script changes, budget changes etc), actor's performance, audience reception and a whole bunch of other elements.

What can be predicted though, based on previous behavior is another provocation attempting to compensate the lack of understanding.

 

[Elsie N]

Digital media may be changing how you think

Digital media may be changing how you think

I was just reading the article in the title and thought there might be parallels to what is being discussed in this thread.

And while the article suggest we all may be having our thinking changed because of our reliance on the digital realm, still the research showed the superiority in the non-digital way of thinking in many instances.

Certainly not a total correlation with the topic at hand, but food for thought.

 

[David Peterson]

As a math grad and geek this was a fascinating read! :-D Thanks!

Sometimes, you have to create a trend by bucking the trend.

How do you buck a trend? By first identifying the trend.... and how can you do that? By using tools like this!

 

[Freya]

I thought it was the dog food arm of the company that took it all down?

 

[Terry VerHaar]

As a math grad and geek this was a fascinating read! :-D Thanks!



How do you buck a trend? By first identifying the trend.... and how can you do that? By using tools like this!

Agree, David. Good luck convincing the Luddites, though. LOL

 

[David Mullen ASC]

I tend to agree more with Marc's analysis -- history has shown over and over again the limitations of these algorithms and yet some people keep believing that next time will be different because the algorithm will be better, and yet there is no evidence that this is true, it's almost a religious belief.

Formulas tend to work up to a limit and then they don't work, because some shift has occurred -- and in the case of movies, the shift is partly due to audiences wanting something different after awhile. A formula doesn't predict when the buyer will be tired of the formula.

This has nothing to do with me being a Luddite, just a cynic -- there's a difference!

 

[Terry VerHaar]

Hah - David I do, of course, respect your perspective but disagree quite a bit. I actually question whether people are reading the same article that I did.

There is a lot of discussion about the fact that there are too many subjective variables to submit them to quantitative analysis. But, unless I completely misunderstood, what is being measured and evaluated is actual human behavior - i.e. what movies people chose to see, what the demographics of those people were (it isn't subjective as to whether you are in Dallas or NYC, male or female), and certain known facts about the films - like who was in the cast. Now, it might be somewhat subjective when you start discussing the genre of the film but I don't think that is a deal killer.

In the article, the author comes right out and says "all models are wrong, but some are useful..." which I think is the point. It should be said that this type of tool is not (IMO) going to tell you how to make your script, except perhaps to give you some thoughts about "bankable" casting. But it is very likely to be helpful deciding what the market potential is for a particular script, how it might be marketed and, as a consequence, something about the size of the budget - at least in the indie world where budget does matter - given it's potential market penetration.

If anyone thinks this is a "push button" model for green lighting scripts or, worse, rewriting scripts to be "marketable" I think they are severely mistaken. But, as another tool to be used judiciously, to inform oneself about how movie-goers are behaving and what they are reacting to, I see genuine potential. The failure of Relativity does illustrate that any model is not infallible. However, I don't believe it negates the potential for such a tool, as the author describes in the article.

BTW - clicking through to the WSJ article within the NFS text is worth the time.

 

[Elsie N]

As a math grad and geek this was a fascinating read! :-D Thanks!



How do you buck a trend? By first identifying the trend.... and how can you do that? By using tools like this!

And how do you identify a trend change? By losing money following the old trend until you've identified the new trend.

 

[David Mullen ASC]

The only truth that seems to be confirmed over and over again is William Goldman's quote "Nobody knows anything" regarding the prediction of box office success in Hollywood. Sure, marketing decisions are based on data, I don't have a problem with that.

 

[Scott Crawley]

I don't entirely disagree with you Terry. I don't doubt that it is possible to generalize about what the average person likes and pander to that lowest common denominator but if Hollywood has shown us anything it is that this approach is responsible for some very average and boring films.

Formula=Predictable. It's just the thing for accounting but the antithesis of a great story. IMHO When you serve up the same film 100's of times, each under the guise of "new and exciting" people eventually refuse to show up and the model fails.

It is Spring so I'll use a baseball analogy since the game is full of stats and I think they bear out the point (though I can't cite any).

Compare batting styles. Batters with a measured swing get on base more often while the ones who swing for the fence every time tend to fly or strike out more often. Which batter drives in more runs? Which one is most likely to make the big hit that everyone remembers? Who does the team look to in the clutch to win the game?

Go big or go home. That is what the Hollywood blockbuster is all about. Why then would you play the bunt?

If you bunt every time you go to the plate they will throw you out every time.

 

[David Mullen ASC]

I think in terms of why Hollywood doesn't make small movies, and their smaller returns, is that there is not a lot of incentive -- agency fees, studio overheads, etc. all prefer getting a slice of larger budgets. I also think there is a franchise mentality at work, again, to reduce variables and the unknown -- every studio wants a hit series so they have less risk because they are more sure of their returns compared to bankrolling an unknown project.

 

[Terry VerHaar]

Scott- Thanks for the vote of confidence. I think. LOL

Actually, you make my point better than I do, by using baseball as an analogy. Baseball is one of the more "big data driven" opportunities in business and, bringing it back around to movies (and books), my recollection of the story behind Moneyball by Michael Lewis (a true story) is just that. Baseball is famous for data, especially the modern game, and Lewis tells the story of how the Oakland A's beat the odds by using it wisely. So, even with the incredible amount of variation and unpredictability in every aspect of the game - from the distance to the fences to the many variations of how a pitcher delivers the ball, patterns emerge and can be acted upon. There is a clear answer to your question about "who the team (should) look to in the clutch.." and it is in the data.

I won't keep beating this horse - which is clearly dead to most of you. However, a few last thoughts -

I am not a Hollywood insider and never hope to be. But by my observation, the Hollywood model may very well be somewhat impervious - and resistant - to the need for a tool like this simply because of the way I understand HW films to be brought forth. HW films are designed to be for "the masses," for "everyone." The simplistic model, as I have heard it told, is that you plan on spending $60 million for domestic marketing and $60 million for foreign marketing, making production budgets almost an afterthought - at least when they are under $40-50 million. With this kind of spending, you may not be able to take a nuanced approach to which cities or which demographics are likely to buy tickets. You tend not to make movies that are meant to appeal to targeted groups.

Indie films may be where something like this will provide more assistance, simply because indie films can't afford to try to appeal to "everyone." Decisions have to be made about marketing, which focus on certain segments of the market. This is, IMO, where a tool can help.

Again, I agree that no one should be writing their script to "pander" - that is, to me, both emotionally and intellectually objectionable. Nor should there be a "formula" for plot development (although the many books on the subject may call that into question), at least from a statistical/data perspective. Where I think a tool like this begins to help is in evaluating where one can find a market for the script/movie that is being made. Maybe it can help with casting decisions (as is pointed out in the article) and maybe the past successes and failures of movies that are "similar" can inform qualitative marketing decisions. As the article's author points out clearly, building models is both a science and an art; it's a "whole brain" thing. As such, it isn't perfect; but neither is decision making of any type when it comes to looking into the future.

As for William Goldman's quote of more than 30 years ago -

“Nobody knows anything...... Not one person in the entire motion picture field knows for a certainty what's going to work. Every time out it's a guess and, if you're lucky, an educated one.”

The idea that one shrugs and gives up when it comes to solving a problem goes against my grain. I think most problem situations, even if not "solvable," can be improved. He is certainly right that there is no "certainty" - that hasn't changed. But, in 1983, the concept of "big data" analysis also didn't exist and a tool such as this would have been unthinkable. Times change! ;-)

Thanks for listening. Over and out.

 

[Elsie N]

Scott- Thanks for the vote of confidence. I think. LOL

Actually, you make my point better than I do, by using baseball as an analogy. Baseball is one of the more "big data driven" opportunities in business and, bringing it back around to movies (and books), my recollection of the story behind Moneyball by Michael Lewis (a true story) is just that. Baseball is famous for data, especially the modern game, and Lewis tells the story of how the Oakland A's beat the odds by using it wisely. So, even with the incredible amount of variation and unpredictability in every aspect of the game - from the distance to the fences to the many variations of how a pitcher delivers the ball, patterns emerge and can be acted upon. There is a clear answer to your question about "who the team (should) look to in the clutch.." and it is in the data.

I won't keep beating this horse - which is clearly dead to most of you. However, a few last thoughts -

I am not a Hollywood insider and never hope to be. But by my observation, the Hollywood model may very well be somewhat impervious - and resistant - to the need for a tool like this simply because of the way I understand HW films to be brought forth. HW films are designed to be for "the masses," for "everyone." The simplistic model, as I have heard it told, is that you plan on spending $60 million for domestic marketing and $60 million for foreign marketing, making production budgets almost an afterthought - at least when they are under $40-50 million. With this kind of spending, you may not be able to take a nuanced approach to which cities or which demographics are likely to buy tickets. You tend not to make movies that are meant to appeal to targeted groups.

Indie films may be where something like this will provide more assistance, simply because indie films can't afford to try to appeal to "everyone." Decisions have to be made about marketing, which focus on certain segments of the market. This is, IMO, where a tool can help.

Again, I agree that no one should be writing their script to "pander" - that is, to me, both emotionally and intellectually objectionable. Nor should there be a "formula" for plot development (although the many books on the subject may call that into question), at least from a statistical/data perspective. Where I think a tool like this begins to help is in evaluating where one can find a market for the script/movie that is being made. Maybe it can help with casting decisions (as is pointed out in the article) and maybe the past successes and failures of movies that are "similar" can inform qualitative marketing decisions. As the article's author points out clearly, building models is both a science and an art; it's a "whole brain" thing. As such, it isn't perfect; but neither is decision making of any type when it comes to looking into the future.

As for William Goldman's quote of more than 30 years ago -

“Nobody knows anything...... Not one person in the entire motion picture field knows for a certainty what's going to work. Every time out it's a guess and, if you're lucky, an educated one.”

The idea that one shrugs and gives up when it comes to solving a problem goes against my grain. I think most problem situations, even if not "solvable," can be improved. He is certainly right that there is no "certainty" - that hasn't changed. But, in 1983, the concept of "big data" analysis also didn't exist and a tool such as this would have been unthinkable. Times change! ;-)

Thanks for listening. Over and out.

Terry, just to be clear, it's not us Luddites vs Terry. '-) That is, this discussion is not zero sum.

In re: your baseball points... How's that working for the A's now? I don't keep up with them that closely but I'm not aware of any hidden gems they've uncovered lately. And Houston, one of the most heavily invested clubs in statistical analysis did very well last year... this year, not so great. They are so statistically minded they even measure the number of rotations each pitcher's pitches make. Presumably by adjusting that it made a better pitcher out of Colin McHue. This year it seems the hitters are catching up.

All I (for one) am saying is that the human decision-making process is probably more important than the data, although data can offer guidance. For me, when writing a script I identify my audience and try to write to that audience. I don't look at data to decide what that audience is... I rely on my life experience. Sometimes that tells me I may not be able to reach a certain audience segment. In some cases, I may not be able to relate as an MIT grad... in others I may not be able to dumb-down enough. '-) Those are human decisions made at the very beginning of the Indie movie-making process like mine that is approached holistically.

Not saying that my process projects as success... not enough data. '-)

 

[Scott Crawley]

I guess the point on which we seem to agree is that the data alone is not the be all, end all. Like anything else we may use the data is just a tool. The art is in the interpretation and application.

 

[Terry VerHaar]

OK - even though I said I wouldn't... lol

... this discussion is not zero sum.

I agree.

In re: your baseball points... How's that working for the A's now? I don't keep up with them that closely but I'm not aware of any hidden gems they've uncovered lately. And Houston, one of the most heavily invested clubs in statistical analysis did very well last year... this year, not so great. They are so statistically minded they even measure the number of rotations each pitcher's pitches make. Presumably by adjusting that it made a better pitcher out of Colin McHue. This year it seems the hitters are catching up.

To be clear - as I stated very much earlier in this discussion, anecdotal evidence isn't going to inform the quantitative approach - and, there will always be outliers and that's good thing so models can be evaluated and tweaks and evolved. No one single film is going to be "guaranteed" success as a result of this analysis any more than one baseball team is going to be guaranteed success in a given season. I really don't know how the A's evolved after the period discussed in the book. I think egos in baseball, like egos in Hollywood, sometimes get in the way of a disciplined approach.

All I (for one) am saying is that the human decision-making process is probably more important than the data, although data can offer guidance. For me, when writing a script I identify my audience and try to write to that audience. I don't look at data to decide what that audience is... I rely on my life experience. Sometimes that tells me I may not be able to reach a certain audience segment. In some cases, I may not be able to relate as an MIT grad... in others I may not be able to dumb-down enough. '-) Those are human decisions made at the very beginning of the Indie movie-making process like mine that is approached holistically.

Not saying that my process projects as success... not enough data. '-)

And to reiterate again (redundancy for emphasis!) - this is NOT a tool for guiding the development of a single script. It might aid in how a story may be tweaked and evolved to enhance success but, more powerfully, it's a tool to guide the marketing efforts to try and bring it to a welcoming audience with the a manageable expenditure.

 

[Terry VerHaar]

I guess the point on which we seem to agree is that the data alone is not the be all, end all. Like anything else we may use the data is just a tool. The art is in the interpretation and application.

Yep!

 

[Marc Wielage]

I think in terms of why Hollywood doesn't make small movies, and their smaller returns, is that there is not a lot of incentive -- agency fees, studio overheads, etc. all prefer getting a slice of larger budgets. I also think there is a franchise mentality at work, again, to reduce variables and the unknown -- every studio wants a hit series so they have less risk because they are more sure of their returns compared to bankrolling an unknown project.

The book I've used as my bible is Lynda Obst's 2014 book Sleepless in Hollywood: Tales from the New Abnormal in the Movie Business, where she talks quite a bit about the death of "mid-budget" pictures between $10M and $60M. One reason is that the marketing costs on (say) opening a $20M film in a thousand theaters works out to much more than $20 million. It makes more financial sense for the studios to gamble on much more costly pictures and then spend $60M-$70M to market them.

Speaking of formula pictures: it's sobering to reflect that Batman vs. Superman made $868M but is considered to be somewhat of a disappointment because it didn't make $1 billion. One reason is that it cost $250M but also cost another $150M for marketing:

http://www.cinemablend.com/new/Batm...ly-Make-Less-Money-Than-Man-Steel-123717.html

Quite a few of the industry wags are wondering how long it will be before one studio gets toppled by making one too many breakeven or money-losing pictures, particularly films that follow the usual summer tentpole/blockbuster formulas. We could eventually see a replay of the crash caused by Heaven's Gate at UA. Already this year, we have five major comic book movies -- Deadpool, Captain America, X-Men, Suicide Squad, and Dr. Strange -- and four major animated films. Not all of these are going to be hits, even though they follow essentially the same formulas and strategies.

Me personally, I'd rather see the studios make ten $50 million dollar pictures instead of two $250 million dollar pictures, but I get that the former doesn't have the potential to make $2 billion as much as the latter. Obst's book points out that this is one reason why a lot of indie filmmakers, writers, and producers are fleeing that world and moving over to cable TV, where they have reasonable budgets and maybe 10-12 episodes to get an audience of one or two million people, which is possible with a decent show.

 

[David Peterson]

history has shown over and over again

Looking at history to help predict the future and guide your current actions is exactly what these tools are designed to do!

**Everybody** does this all the time, every day of their life. Just some of us do it via only informal means, others use formal methods as well.

All we're doing is debating the ratio of formal vs informal. I'm a strong supporter of increasing the level of formal methods, as it lends more rigor, accuracy, and structure to the process.

Of course you shouldn't just blindly follow the results, human knowledge should still be used to interpret the computer results and then determine the actions to take next from it.