5 ms·

That's the mass taste. I am not trying to discuss with the mainstream taste, just made the tool. For those interested in getting a personalized movie ranking I

by arekp 14y ago

That's the mass taste. I am not trying to discuss with the mainstream taste, just made the tool. For those interested in getting a personalized movie ranking I created another tool "Movie Galaxy" http://arek-paterek.com/movie-galaxy/ http://arek-paterek.com/movie-galaxy/

nollidge 14y ago
> That's the mass taste. I seriously doubt that. Just because it's what the available data show doesn't mean that that's the popular taste. If you surveyed 100 people who have seen both The Avengers and Godfather Part 2, I highly doubt a majority will regard the former as a better film than the latter. People are in different moods when they rate films, older films don't get rated as often, the scale with which an individual rates different films may be wildly non-normalized... Website idea[0]: show two random movies, ask them if they've seen both, and if they have, ask the user which one's better without prescribing what "better" means. Repeat 1 trillion times. I'd be much more willing to trust differential data like this. [0] - not going to call it a startup idea, because I have no idea how it would make money.
- tatsuke95 14y ago
  >"I seriously doubt that." I think you'd be surprised. Though, I also think you'd be hard-pressed to find people who have watched both films (I've only seen one, and it isn't The Avengers). And there's a reason for that: most people don't want to watch a three-hour, epic drama from 1972. Ask people who haven't seen either film which they'd rather see, tonight, and I'll bet they choose The Avengers. Why? Because if you haven't seen GF2 by now, it probably doesn't interest you. So on some level, these rankings speak to the current Zeitgeist. I can't understand the vitriol directed at these rankings. Sure, there is a definite bias towards modern films in the imdb rankings, but it is what it is. I don't think that it's any different than some crotchety old movie critic telling me which films I should like best. The fact of the matter is, there is no objective measure of "best". Scan these comments. I can find examples to agree with and disagree with in nearly every single one. But they are all saying the same thing: "This list is wrong!" Based on that, I don't think you or I would have any more success curating a list of movies. This list is as good as any. Anyway, great tool.
- scottjackson 14y ago
  > Website idea: show two random movies, ask them if they've seen both, and if they have, ask the user which one's better without prescribing what "better" means. Repeat 1 trillion times. http://flickchart.com http://flickchart.com
  - pbhjpbhj 14y ago
    That's really neat. It's quite addictive to rank films that way. Annoyingly hard in some instances. The only thing I think is lacking is some way to get a reminder about the plot; when you're talking about movies seen more than a couple of decades ago ... A "would like to see" button would be useful too.
  - hansy 14y ago
    Would the two movies be from the same genre? For example, I would have a difficult time deciding which of the two - "The Secret in Their Eyes" or "Gladiator" - is better.
    - dagw 14y ago
      Having played with the site for the past 20 minutes now, the by far most difficult thing is choosing between two horrible movies. I really wish there was a "they're both equally horrible" button.
binarysolo 14y ago
Out of curiosity, how'd you do feature selection for the clustering (on a high level) for Movie Galaxy?
- arekp 14y ago
  The clustering was based on the regularized SVD / matrix factorization results. Details in my e-book.
allforJesse 14y ago
And interestingly, "Research suggests that the best way to predict how much we will enjoy an experience is to see how much someone else enjoyed it." - “If money doesn’t make you happy, then you probably aren’t spending it right” by Elizabeth W. Dunn, Daniel T. Gilbert
Steko 14y ago
Maybe put in an inflation adjustment factor. That would reflect how a movie rates compared to it's contemporaries. You'd have to be careful not to cap a movies' rating just because scores in that era are more lenient though -- maybe pull movies toward an overall median based on their distance from a 2-3 year moving average. So say we want the median score to be 50%. And the median rating in 2008 was 63%. Let's take two movies from 2008 -- one scored 98%, the highest score in several years, and one scored 63%. The 98% movie stays at 98%. The 63% movie score becomes 50%.
ekianjo 14y ago
The tool has no point if the data source is flawed. It is like saying you made a weather forecast website but none of the predictions are accurate: It becomes worthless. If you make a tool that tool has to be able to extract quality data before anything else.

by arekp 14y ago

nollidge 14y ago

> That's the mass taste. I seriously doubt that. Just because it's what the available data show doesn't mean that that's the popular taste. If you surveyed 100 people who have seen both The Avengers and Godfather Part 2, I highly doubt a majority will regard the former as a better film than the latter. People are in different moods when they rate films, older films don't get rated as often, the scale with which an individual rates different films may be wildly non-normalized... Website idea[0]: show two random movies, ask them if they've seen both, and if they have, ask the user which one's better without prescribing what "better" means. Repeat 1 trillion times. I'd be much more willing to trust differential data like this. [0] - not going to call it a startup idea, because I have no idea how it would make money.

tatsuke95 14y ago
>"I seriously doubt that." I think you'd be surprised. Though, I also think you'd be hard-pressed to find people who have watched both films (I've only seen one, and it isn't The Avengers). And there's a reason for that: most people don't want to watch a three-hour, epic drama from 1972. Ask people who haven't seen either film which they'd rather see, tonight, and I'll bet they choose The Avengers. Why? Because if you haven't seen GF2 by now, it probably doesn't interest you. So on some level, these rankings speak to the current Zeitgeist. I can't understand the vitriol directed at these rankings. Sure, there is a definite bias towards modern films in the imdb rankings, but it is what it is. I don't think that it's any different than some crotchety old movie critic telling me which films I should like best. The fact of the matter is, there is no objective measure of "best". Scan these comments. I can find examples to agree with and disagree with in nearly every single one. But they are all saying the same thing: "This list is wrong!" Based on that, I don't think you or I would have any more success curating a list of movies. This list is as good as any. Anyway, great tool.
scottjackson 14y ago
> Website idea: show two random movies, ask them if they've seen both, and if they have, ask the user which one's better without prescribing what "better" means. Repeat 1 trillion times. http://flickchart.com http://flickchart.com
- pbhjpbhj 14y ago
  That's really neat. It's quite addictive to rank films that way. Annoyingly hard in some instances. The only thing I think is lacking is some way to get a reminder about the plot; when you're talking about movies seen more than a couple of decades ago ... A "would like to see" button would be useful too.
- hansy 14y ago
  Would the two movies be from the same genre? For example, I would have a difficult time deciding which of the two - "The Secret in Their Eyes" or "Gladiator" - is better.
  - dagw 14y ago
    Having played with the site for the past 20 minutes now, the by far most difficult thing is choosing between two horrible movies. I really wish there was a "they're both equally horrible" button.

binarysolo 14y ago

Out of curiosity, how'd you do feature selection for the clustering (on a high level) for Movie Galaxy?

arekp 14y ago
The clustering was based on the regularized SVD / matrix factorization results. Details in my e-book.

allforJesse 14y ago

And interestingly, "Research suggests that the best way to predict how much we will enjoy an experience is to see how much someone else enjoyed it." - “If money doesn’t make you happy, then you probably aren’t spending it right” by Elizabeth W. Dunn, Daniel T. Gilbert

Steko 14y ago

Maybe put in an inflation adjustment factor. That would reflect how a movie rates compared to it's contemporaries. You'd have to be careful not to cap a movies' rating just because scores in that era are more lenient though -- maybe pull movies toward an overall median based on their distance from a 2-3 year moving average. So say we want the median score to be 50%. And the median rating in 2008 was 63%. Let's take two movies from 2008 -- one scored 98%, the highest score in several years, and one scored 63%. The 98% movie stays at 98%. The 63% movie score becomes 50%.

ekianjo 14y ago

The tool has no point if the data source is flawed. It is like saying you made a weather forecast website but none of the predictions are accurate: It becomes worthless. If you make a tool that tool has to be able to extract quality data before anything else.