Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

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

This was a big story a few weeks ago. I think the bottom line is the old adage "nobody knows anything" particularly applies to Hollywood and hit films. I think they're just as perplexed when a movie unexpectedly bombs as when it unexpectedly becomes a hit. Too many wild cards, all kinds of chaos theory.
 
I think that is the whole point behind "big data" and gleaning more out of information than humans alone can recognize. The decisions that go into spending big dollars (or even small dollars) are rarely "rational" in and of themselves, as it is now. If you have any knowledge at all about what deep quantitative analysis has done in the world of investments and investment theory, then you might not be so quick to dismiss it. It may not be pure science but it beats pulling your decision out of thin air - or your ass.
 
Terry VerHaar said:
I think that is the whole point behind "big data" and gleaning more out of information than humans alone can recognize. The decisions that go into spending big dollars (or even small dollars) are rarely "rational" in and of themselves, as it is now. If you have any knowledge at all about what deep quantitative analysis has done in the world of investments and investment theory, then you might not be so quick to dismiss it. It may not be pure science but it beats pulling your decision out of thin air - or your ass.
Click to expand...

The approach ignores the unpredictable and unquantifiable elements, many of which exist in creative domain and make or break it, and in this context it is complete nonsense.

It was doomed to failure from the very start and indicted by its own creators, blinded by figures, clueless about what the creative domain consists of, relying on partial perception of reality and arrogantly convinced they have gotten it all figured out.
 
Terry VerHaar said:
I think that is the whole point behind "big data" and gleaning more out of information than humans alone can recognize. The decisions that go into spending big dollars (or even small dollars) are rarely "rational" in and of themselves, as it is now. If you have any knowledge at all about what deep quantitative analysis has done in the world of investments and investment theory, then you might not be so quick to dismiss it. It may not be pure science but it beats pulling your decision out of thin air - or your ass.
Click to expand...

It is an interesting idea that could improve the bottom line a little but I think it fails to address the real problem; Risk aversion makes for heartless drama and weak-assed, no suspense thrillers.
 
I still have to disagree with the idea that this approach cannot be a valuable tool. No one says it takes over the decision-making anymore than a RED shoots a beautiful movie all by itself. The things that a set of algorithms can be set up to look at are most likely the same things that a human will look at when they are deciding whether to green light a script. The big data analysis just does it more thoroughly and more objectively; and a whole hell of a lot more quickly.

Hrvoje -I agree that a truly "breakout" film, in terms of it's plot elements, casting approach and any other unique aspects, would probably not fare well in this analysis - just the way a "black swan" event in the investing world is never predictable - except in the inability to predict it. So a really unique film would get ignored. But how many films like that will make it through any "typical" Hollywood selection process when big dollars are at stake? From my observation, Hollywood rarely has that kind of courage.

Scott - If you mean it would "improve the bottom line a little bit" on a single film, I doubt that. In the end, I think the effect of an accurate approach (admittedly a high bar) would be to provide substantively better returns over a portfolio of films. I doubt you could count on it making a single film better simply because I don't think the analysis is designed to help one alter an existing script to be more appealing to an audience. Personally, I do think that would be a hopeless strategy.

As the article points out, the assumptions are critical to the success of the approach. And so, unless and until the questions being addressed are the right questions to ask, the model will suffer from the phenomenon of garbage-in, garbage out. I fall on the side of trying to refine an approach like this as a useful and potentially valuable tool. As a Hollywood outsider, my observation is that the industry isn't doing all that well as it is. To overstate the case a bit, should they just continue to give a monkey a typewriter and produce whatever script comes out, hoping one of the them is Gone With the Wind? What do they have to lose - except maybe the occasional bankrupt studio - of course THAT never happens under the "usual" approach.

Peace.
 
Last edited:
Terry VerHaar said:
Hrvoje -I agree that a truly "breakout" film, in terms of it's plot elements, casting approach and any other unique aspects, would probably not fare well in this analysis - just the way a "black swan" event in the investing world is never predictable - except in the inability to predict it. So a really unique film would get ignored. But how many films like that will make it through any "typical" Hollywood selection process when big dollars are at stake? From my observation, Hollywood rarely has that kind of courage.
Click to expand...

Creative domain is by its very nature changing, impulsive and unpredictable throughout its large part.
It consists of many elements residing out of analytical domain. Elements which cannot be quantized and put into an algorithm.

Some didn't get the memo so someones' pile of imaginary value got squandered to pay the price for extreme case of stupidity.
 
the amazing thing about this story is that people who were investors actually stayed with it. The original guy who started this model must be one hell of a salesman.
 
Hrvoje Simic said:
Creative domain is by its very nature changing, impulsive and unpredictable throughout its large part.
It consists of many elements residing out of analytical domain. Elements which cannot be quantized and put into an algorithm.

Some didn't get the memo so someones' pile of imaginary value got squandered to pay the price for extreme case of stupidity.
Click to expand...

I think your assessment of Hollywood movie making as being a "creative domain" that is "changing, impulsive and unpredictable" is far too kind. LOL
 
Terry VerHaar said:
I still have to disagree with the idea that this approach cannot be a valuable tool. No one says it takes over the decision-making anymore than a RED shoots a beautiful movie all by itself. The things that a set of algorithms can be set up to look at are most likely the same things that a human will look at when they are deciding whether to green light a script. The big data analysis just does it more thoroughly and more objectively; and a whole hell of a lot more quickly.
Click to expand...
The problem with trying to apply past financial data models to filmmaking is that things constantly change, particularly modern audiences. What works 5 or 7 years ago may not work today. A good example is Dreamworks Animation, which had 4 or 5 major flops in a row, ultimately resulting in CEO Jeffrey Katzenberg selling the company to Universal this week. They had a fairly decent 20 year run, but eventually the old formula no longer worked. And there's often an indefinable "X Factor" in what makes one film connect with audiences, and another film pushes them away.

It's also a problem for studios trying to figure out hits. When the filmmakers behind Deadpool made that film last year, Fox was unconvinced an R-rated smart-ass superhero film could do well, and they chopped the budget down to $60M and figured they'd eventually make a small profit. Everybody was blown away when it wound up making almost $767 million dollars, the biggest R-rated film in history. The distribution execs were happy, but also complained that they had no data on any kind of film like this because nobody had ever done one before. They were glad to make the money, but embarrassed that their highest estimate was more than 50% lower than what the film ultimately made.

Many, many industry pundits have predicted that it's only a matter of time before one studio puts out too many bomb tentpole releases in a row, which could potentially topple the studio. Don't believe it can happen? It did back in 1980, when Heaven's Gate essentially bankrupted United Artists. Just as some believed the Wall Street banks were "too big to fail," the danger is that nobody is too big to fail, given a perfect storm of misfortune and bad decisions.
 
Anecdotal stories illustrate nothing; except perhaps what you already believe to be true. And, that we all love to tell and hear stories. I once made 10x my money in a stock. I once lost my whole investment in the stock of a company that went bankrupt. What does that say about my investment capabilities? Nothing.

"Things constantly change" in the investment world as well; international economies, fiscal and monetary policies, business results, people's attitudes. Hell, world weather patterns change a lot of things for business people. Economics is not the study of money but the study of people's behavior. So, to say that there is nothing to be gleaned by studying the ongoing aggregate behavior of people in a certain area of business or economics is simply defeatist. I'll reiterate, that's exactly what business people (in any business) attempt to do whenever they make business/product decisions. They often do it with large amounts of data analysis available to them, and they often do it "seat-of-the-pants." (Donald Trump would seem to be a good example of the latter, according to him. LOL) Personally, I'll take information over "flying blind." YMMV

And I will reiterate, analysis of past patterns is a) only a tool to be used wisely, and b) is never going to predict the anomaly of a "black swan" event, by definition. IMO, only an independent filmmaker is likely to produce that "outlier," anyway.

I am sure I am not going to change anyone's mind who doesn't "believe in" the value of disciplined data analysis, which is fine. But I still wonder exactly how people expect to make business decisions, even in a so-called "creative" business. Years of applied experience, good judgment, clear and objective thinking, intuition? These are all very helpful - even essential - to the process, for sure. But tools can matter Even if I have Roger Deakins as my DP, I'd still rather give him a Weapon or an Alexa instead of an iPhone when I ask him to shoot my movie.

And just as an aside - I think the idea that banks are "too big to fail" is not meant to suggest that they can't fail (as we have seen). I believe the point made by that statement is that their failure is so consequential to the world economy that they are actually too big to let them fail. Hence, the fear around letting them get that big is that we have to expend massive resources to rescue them if/when they do.
 
Improved Computing Cycles

Improved Computing Cycles

I think we are closer to a new frontier using data model simulators more successfully outside of strictly scientific applications. If a model fails, its probably more related to information gaps or input variables being too thinly sliced. With improved computing cycles, there will or should be exponential growth with recursive self-improvement algorithms.

Business and finance have been more elusive, because there are metaphysical components involved in what we generally classify as successful or profitable.

The studio in discussion, probably was just a victim of being ahead of its time using this formula or a variant of it exclusively, along with any poor analog decision making. But, these 'moneyball' models will become increasingly important as our earthly resources vanish, government and policy bodies will fuel the need for more physically accurate models not just fictitious simulations.
 
M.Halsell said:
I think we are closer to a new frontier using data model simulators more successfully outside of strictly scientific applications. If a model fails, its probably more related to information gaps or input variables being too thinly sliced. With improved computing cycles, there will or should be exponential growth with recursive self-improvement algorithms.
Click to expand...
As long as movies are primarily about people and emotions and feelings, I don't think this can happen. The danger of using algorithms to decide what movies are made is that a) you wind up making a bunch of movies that are very much like each other, and b) you don't wind up making movies that break from the norm and try something radically different.

There's always a case you can make for gut instinct, where you come up with a story and characters and make a film because it's a story you want to tell -- not necessarily because it will make money, and not necessarily because anybody has ever done it before.

Don't forget: 7 Hollywood studios turned down George Lucas' script to Star Wars (Universal's Ned Tanen infamously turned it down twice), and it was only happenstance that exec Alan Ladd, Jr. at Fox took a chance and decided to make it despite all the financial and marketing people advising him not to...

http://www.forbes.com/sites/judebre...ovie-mogul-youve-never-heard-of/#1d65413e3734

Sometimes, you have to create a trend by bucking the trend. There's an interesting book on this subject, George Lucas's Blockbusting: A Decade-by-Decade Survey of Timeless Movies Including Untold Secrets of Their Financial and Cultural Success. It's a wordy title, but it's a fascinating 976-page book covering every successful film of the last 90 years and recounting what the films cost to make, how they were made, why they were made, and venturing good opinions on why they made money. Some of the stories that were told on some of the major bombs of the past are sobering, particularly when at the time they were "can't miss" ideas that the studios were absolutely sure would make big money, since they fit all their bullet points for what was necessary in a commercial hit.
 
The very nature of the subject matter leaves ample room for debate. Admittedly, I'm no where near the Hollywood machinery, so feel free to correct my logic.

The film industry by all accounts is a risky business financially. The majority of films are unprofitable with just a handful of successful films anchoring industry profits. Protecting the creative process could easily by definition be described as risk aversion.

Just looking at Relativity Media's Portfolio, Box Office totals against Production Costs, very few of the titles were underperforming. I don't think the model was completely flawed. Without the whole backstory, it's difficult to concretely form opinion. Most industries have weaponized data modeling. It hard to accept that the roundtable 7 aren't agressively looking for an improved formulaic approach.

But I gotta say, the Hollywood return rates do resemble a sophisticated game of Cee-Lo.
 
Every business that sells anything is about "people and emotions." Does anyone believe that soap or cell phones or cameras are sold based on simple pure logic? Are products designed based on functionality alone?. As I said earlier, economics is not the study of money or business, it's the study of behavior. This is no different. "Gut instinct" is often decision making without all the facts. Is there place for judgment? Yes - and the better informed it is, the more likely it will lead to success.

And then there is the occasional wild-ass-guess. Sometimes that works out - but would you bet your kids' life on it?
 
Searching for a fixed model is wandering through a labyrinth of illusion.
There is no fixed model for a changing realm.

Data analysis analyzes what is already existing.
There is no data to analyze of something yet to happen.

Data analysis covers that which can be analyzed.
There is no analysis of that which cannot be analyzed.

Shortcuts avoid paths necessary for development.
There is no success which is not based on development.
 
A couple of thoughts...

I would hardly call analysis of large amounts of dynamically changing information on an ongoing basis, "fixed." Nor is anything "fixed" when the assumptions you apply and the data they uncover are constantly in flux.

Also let's remember that analysis like this is essentially looking at human behavior and, whether we like to acknowledge or not, patterns of human behavior - in the aggregate - are all too predictable. If rational decision making is not influenced by the world we observe, then what is it based upon?

When it comes to the supposed inability to use past behavior to predict future occurrences, I'd ask - what other means do you employ? How does one avoid any influence from past occurrences when they make decisions? Is there a suggestion here that humans making decisions are NOT influenced by the past? Exactly what methodology - including even "set-of-the-pants" decision making - actually casts off the idea of using past patterns of human behavior in making decisions?

As for this...

"Data analysis covers that which can be analyzed.
There is no analysis of that which cannot be analyzed."

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

Over and out...
 
I agree with Terry on this one. It's technically doable. If a buddy asks for money and doesn't pay it back twice in a row, a third event is corrected behavior based on the past analysis.

I also feel the notion of the Hollywood cornerstones as the oracles of creativity, largely are based on having landed on Plymouth Rock first. I actually have more deference to the new content players that are carving out markets in highly fragmented, competitive environments.
 
I guess polling is data-gathering for analysis. But when I am polled, I am thinking they aren't asking the right questions.

I. e, "Are you better off now than you were 10 years ago?"

Answer: Yes and no. I am better off health wise but not as well off financially. I am happier but I travel less.

My point is, I think statistical analysis can be a helpful tool in a broad sense but not so much when applied individually. So if the industry wants to know how it is doing and perhaps even where it may be heading in the future, data analysis could give them something to plan by... especially if there are multiple outcomes projected based on potential changes in future variables.

But for individual outcomes such as the success of a movie, I just think there are too many variables to give anything other than a hint of future success. We've all seen a movie made and ready to be released and suddenly there is a world event that too closely is associated with the movie, and the movie is either cancelled or is released and is extremely successful due to the association.

And not only that, there are a myriad of personalities involved in movie-making that may affect the finished product. And while the story may or may not resonate, how the director interprets it or how the look for the picture in relationship to the story is presented can impact the success of the finished product. Variables of individual past performance needs to factor into the equation... not just the broad data.

But the thing I remember about systems was advice from a gambling friend of mine. He said once a betting system is identified and becomes widely known, the "house" will move to nullify it. I think the house terminology also applies to trends... that is, trends change as the collective mood changes.
 
Last edited:
You must log in or register to reply here.
Back
Top