The algorithm or art
Recently, I’ve read not one but two articles on using algorithms to predict the success of movies. In both cases, the analysts start with a script (one team uses human analysts, the other a computer), then analyze it for key elements, plot twists, setting, characters, and so on. Then they use what appears to be regression analysis to determine the degree of commonality the script has with other hits.
And it works. Hits show up. Misses are clearly seen.
This isn’t entirely new, of course. There have been a couple of attempts to do this with music, for example. Over at Platinum Blue they use algorithmic analysis of music in order to tell which of the sixty types of hit songs a given piece might be. They can tell you that “sorry, you’re somewhat off on the bass track and only 65% likely to be a hit.” They can’t tell you how to fix it, though.
The case has been made that this doesn’t lead to gross uniformity in the arts. After all, currently publishers and funders of all sorts routinely fail to take a flyer on unproven talent or stuff that seems a bit off the beaten track. But underneath the unproven aspects and the unfamiliar elements might be the skeleton of something completely familiar, something designed, inadvertantly or not, to tap into something deep in the human brain.
Of course, the recipe here is not for artistic success; it’s for commercial success. Those sixty hit song archetypes go through fads, they fade in and out of popularity (Beethoven and Mozart are apparently solidly in one of the clusters). Obviously, some people have this algorithm in their heads: Lennon & McCartney each seem to have had half of it, and people like Billy Joel who manage hits spanning 30 years must have a pretty good knack for it, and legendary producers like Clive Davis maybe even have some of it written down. It’s not about “good” but about “popular.”
This past spring, for instance, he analyzed “Crazy,” by Gnarls Barkley. The computer calculated, first of all, the song’s Hit Grade—that is, how close it was to the center of any of those sixty hit clusters. Its Hit Grade was 755, on a scale where anything above 700 is exceptional. The computer also found that “Crazy” belonged to the same hit cluster as Dido’s “Thank You,” James Blunt’s “You’re Beautiful,” and Ashanti’s “Baby,” as well as older hits like “Let Me Be There,” by Olivia Newton-John, and “One Sweet Day,” by Mariah Carey, so that listeners who liked any of those songs would probably like “Crazy,” too. Finally, the computer gave “Crazy” a Periodicity Grade—which refers to the fact that, at any given time, only twelve to fifteen hit clusters are “active,” because from month to month the particular mathematical patterns that excite music listeners will shift around. “Crazy” ’s periodicity score was 658—which suggested a very good fit with current tastes. The data said, in other words, that “Crazy” was almost certainly going to be huge—and, sure enough, it was.
In The Long Tail, Chris Anderson makes the case that hits are actually less emotionally satisfying that media that targets an individual’s particular likes and dislikes. They push a lot of buttons, but not very well.
You can think of our tastes as being like a box of springs. We all have the same set of springs, but they are tuned differently, some tightly coiled and some not. When a spring is depressed all the way, whatever is resting on it is “hitting our buttons.” A hit is something that lays across most of the springs with some moderate weight. It effectively presses all the buttons, just not very well. A tightly focused niche title aims for just a few springs, and presses down really hard; if you have loose springs there, you’re gonna love it. Everyone else “won’t get it.”
But again, we’re not talking about artistic merit. You see, the springs are for things like harmonic overtones, excitation in specific frequency bands, that sort of thing. They are about mechanics, about execution, and when you reduce the world to clockwork, sometimes the art might fall out. (Yes, I’ve mentioned this before.
The only reason this can be done with movies and with music is because there is enough of an established atomic structure to these media that you can do analysis on it. Even with something that seems as subjective as a story, there’s enough elements that can be isolated that a neural network can chew on them. (Check out the revisions to The Interpreter in Gladwell’s article to see how it works).
Games, however, lack this. We instead have something more akin to the disastrous attempts to make the “ideal painting,” that were done purely subjectively, via opinion polls. And the thing is that people don’t actually know their own buttons very well.
Sixty-seven percent of respondents liked a painting that was large, but not too large — about the size of a dishwasher (options ranged from ‘’paperback book’’ to ‘’full wall’’). A whopping 88 percent favored a landscape, optimally featuring water, a taste echoed by the majority color preferences, blue being No. 1 and green No. 2. Respondents also inclined toward realistic treatment, visible brushstrokes, blended colors, soft curves. They liked the idea of wild animals appearing, as well as people — famous or not — fully clothed and at leisure.
Can games get there? Yes, I think they can, but it’ll take analysis at a deeper level than what we have currently, followed by an exhaustive historical survey of games to build the database and clusters.
And then we’ll get to have the debate about whether we should be doing it at all.
In the meantime, Platinum Blue offers a consumer service at $10 a song, and I am inclined to give it a try.
17 Responses to “The algorithm or art”
Sorry, the comment form is closed at this time.
Original post: The algorithm or art by at Google Blog Search: on it
Over at Raph Koster’s Website, he’s got a post up about “The algorithm or art.” Using feedback to determine the “worthiness,” whether economic or artistic, of songs, movies and games. Back when I was at Cornell and actively involved in the student poetry writing scene there, we had weekly open-mic and other similar events, gang
Raph Koster has a good post up about the algorithmic structure of hit production, and why computer analysis can predict the commercial success or failure of Hollywood movies and pop songs. The explation offered as to why even works which we know are aesthetically sub-par can nonetheless ‘push our buttons’ and evoke strong positive
Raphs Website The algorithm or art
Fascinating. Spock lives and string theory and all that.
Yeah, go for it on the song.
Games probably can do something on this in a basic way right now. The biggest problem is developing what to measure.
MMO’s, I’m thinking, would push the few strings hard. I’m wondering right now if that’s not the case on almost all games.
Overall, this is a cool study on human nature.
Of course not. Our brains are trained to study the world around us, an astounding variable as is, and then able to go far beyond. Then you can add emotions on top of that. The variable are so immense that you’d have to literally be God before you can start hitting on uniformity in this vast sense.
That’s both compelling and terrifying. I can already sense which of my songs have the strongest hit potential, but it sounds like the additional information that they provide would be very interesting. I can write some pretty catchy stuff, but it could easily be before or after its time.
Strange Agency uses proprietary software to “calculate” game genres.
See: Tomb Raider and the Genre of Stealth
Komar & Melamid’s experiments in poll-driven painting were more hilarious than disastrous; toungues were firmly in cheeks throughout the whole process. They also made some poll-driven music, recording the “most wanted” and “least wanted” songs, both of which ultimately fall into the “so bad it’s good again” genre. (http://www.diacenter.org/km/musiccd.html)
For me, the question to an game-appeal-test-o-matic would be this: Is a tool that tells me my game is not a hit — but doesn’t tell me anything about what the problems are or how to fix them — better or worse than a human play tester that tells me they don’t like my game, and provides me with lots of subjective (and probably bogus) advice about what they’d like to see?
The mention of an attempt to make the ideal painting via opinion polls reminds me of this site I came across recently:
http://thefairest.info/
What is the fairest photograph of them all? “TheFairest.info is a project to try to find the prettiest image in the world, using voting and some algorithms.”
Basically a “hotornot” for true beauty. (They’re almost all nature photographs… no pr0n.) I find it interesting; could a site like this find a photograph which was “objectively” the most beautiful in the world, as an aggregate of the “subjective” responses?
Also enjoy the more lighthearted sisters of this site:
http://thefunniest.info/
http://thecutest.info/
And if you haven’t already, enjoy “xkcd”, the great webcomic written by the creator of these sites:
http://xkcd.com/c173.html
http://xkcd.com/c161.html
I think there is an equivalent here for games and ventures like Allen Sligar’s Gamemetrics and the research at Play On (PARC) should help us get there. Richard Bartle once spoke of a student that manually tracked and tagged the forums for feedback and contextual shifts (or something like that). With more Theories of Fun and Gammar of Games, we’ll get there.
While popularity may not be art. There are lasting innovations where popularity and art intersect. The Old Masters of painting were masters at applying new techiques. The ‘old masters’ or Rock ‘n Roll were also masters and innovators. So, I think we can get there.
F
[…] […] Over at Raph Koster’s Website, he’s got a post up about "The algorithm or art." Using feedback to determine the "worthiness," whether economic or artistic, of songs, movies and games. […] […]
Admitedly I had to think about this before responding to the thread, mainly because it touches somewhat closely to what we’re doing. And well we’re pretty close to launch now, barring any more unforseen issues, so it dosnt matter anyhow, because people will understand what we’re doing in a month or so anyway, of course 500k for a marketing budget would help but hey you cant have everything :). PS: for anyone thinking about doing a startup and bootstrapping, I can tell you its very hard.
What you are reffering to is actually called predictive modeling. Its used widly to predict not only price points and costs but trending and modeling for programs, as well as consumer prefferance (a newer application of data mining)
This is used widely in health care, retail, petroleum, manufactured goods, et al. And its used because there are a number of factors in place to allow for it.
1. Uniform (or somewhat) industry launguage to describe your market/industry. (easier for things like medicine, law, geology, music!, marketing, business, and goods you can physically describe not so for art* and design*)
2. A good analysis and inventory of all previous permutations and offerings to the market, (in other words an aggregated data set that examines all the component parts, what platinum blues done for music, or in movies the 36? standard plots and all the movies that have used them, in health care the different types of diabetes, thier stage, and associated co-morbidities and costs, etc.)
Morgan mentioned strange agency, my impression from thier site is that they might have something along these lines, and Frank mentioned the Bartle piece, I believe there was a TN tread on this I commented on, if I recall the proposition from the academics was manual review of the text for acadeic rigor was a good method, I believe I suggested conversion of the forums to XML, setting tags and aggregating it into a database for data mining, I dont think anyone agreed with me 🙂
3. A significant sample size representative of your market. (I wont get started on this again)
4. An ongoing method of capturing and adding to the data set.
5. Representative data points, not just descriptive data, but measurable data. Measures and dimensions that can be used to extrapolate information to help decision making.
Admittedly I’m a noob related to the gaming industry. Theres still a lot for me to research, read and understand but I’ve done my best to understand where the gaming industry is related to 1-5 above. And the damning conclusion I’ve come to is that well….I’m not sure the industry has any of it really, my feeling is that people think: Well why go through all that trouble when we could just be making games? And who finds that kind of thing interesting anyhow?
I find that kind of thing interesting, whats more I like games and I think it might be able to help the “ideal painting” a bit easier to make, or at least cost less, appeal more, and help the industry understand what gamers like, dislike, and think is important.
Now on to the interesting part.
Raph, please finish A Grammer of Games. from what I’ve read it apples to #1 above or goes a long way toward it. I emailed you very excited when I first read it, because I think that its seriously that important, and other works that describe the data that needs to be captured go a long way toward it. What I’m saying is that gaming needs a language, #$%^ that it needs an alphabet too.
When #1 occurs even if its a work in progress, those who’ve done #2 can apply the principles of #1 to set up the foundations for predictive modeling. However #2 is best handled by academics, its subject to much debate and otherwise reasonable people will be disagreeing on it. (note this is definately PhD thesis bait, just make sure its in a SQL DB pls thx 🙂
What we’re doing at GameMarketMetrics is #3-5 above, predictive modeling is possible with only 3-5, but it’ll take time (basically until historical trending can take place) to be accurate without 1-2 above, for various complex reasons that would likely bore the hell out of anyone not into data.
Art* & Design* I’m not an MFA, but I know that art and design can be broken down into components of color, subject, form, function, medium etc. this is what the “language of games” IMO must be able to accomplish.
This does not take the place of the ART of game design, from a strictly mechanist POV data mining and predictive modeling can tell you what people like/dislike and what thier likely to prefer or trend thier consumption, it will not tell you HOW to design the component parts, or what the house should look like, I dont think what constitutes “asthetically pleasing” in a game can be accurately predicted, because art elicits an emotional response, and people are not robots, what I’m saying is there are huge limitations to what data can tell you, when it comes to art and design, and I dont think data can or should take the place of creativity. I do think it can help us understand the subject better.
Especially if its affordable and easy to get to 🙂
PS: I want people to understand that I feel there is a lot of room to develop more of the data set than what we have already, part of this will come from the community (players), but there will also be a space for designers, academics and industry stakeholders to propose new data, or help us refine what were already capturing as well as ask questions and etc. 3-5 alone is a big order, I know I have not thought of everything, we’re in a new space and it’ll take input from a wide range of people to improve our relavence. So basically I’ll be asking for input and hoping people feel what we’re doing is important enough to bother giving us some.
We’ll be at the GDC this year as well and so if anyone is interested in talking more let me know.
Allen
ceo at gamemarketmetrics.net
[…] Raph’s Website » The algorithm or art I’ve read not one but two articles on using algorithms to predict the success of movies. (tags: https://www.raphkoster.com 2006 art algorithms predição) […]
[…] Raph writes a great post called “The algorithm or art“. He also links over to a fascinating art experiment where two Russian painters made “ideal paintings” based on opinion polls — the results being kitsch. There are two things that come up in this discussion: letting mass-thinking lead innovation and trying to discover recurring components of popularity. The former is usually a disaster and the latter is fairly spooky, but oddly intriguing. […]
[…] Raph Koster has a great write-up on a computerized movie spoiler (script summary) analysis tool. The tool predicts commercial success of a given spoiler. I would love to have one of these for games. […]
[…] alecaustin[Tags|computational media analysis]Raph Koster has a good post up about the algorithmic structure of hit production, and why computer analysis can predict the commercial success or failure of Hollywood movies and pop songs. The explation offered as to why even works which we know are aesthetically sub-par can nonetheless ‘push our buttons’ and evoke strong positive reactions are quite persuasive, particularly when niche products are described as pushing a more limited set of buttons really hard. Sounds about right to me… linkpost comment […]
[…] Kawaguchi Dimitri Vorkapich Lansky “Mild und Leise” mp3 Lansky & Radiohead Remko Scha Raph Koster Xah Lee Algorithmic Art Programs Alvy Ray Smith PostModernEssay Generator Clifford A […]