| Author | Thread |
|
|
08/02/2004 02:43:46 AM · #1 |
I have no clue how the server decides what order to give you the photos for voting but it sure isn't random. I can't count how many times I will see 3 or 4 images grouped together because they both have the same word in the title, or are nearly identical in subject matter. And no I'm not talking about seeing 4 insects in a row, I'm talking about very specific things that would almost have to be manually grouped. I was told they are random in delivery, sure doesn't seem that way.
|
|
|
|
08/02/2004 02:55:10 AM · #2 |
| I was annoyed with the "randomness" today myself. I skipped one photo since I didn't know what the hell it was, and as I was voting it came up every other photo. Grrr. |
|
|
|
08/02/2004 07:27:34 AM · #3 |
It is random. Everyone sees the images in a different order -- so the particular anomalies your seeing are unique to your grouping. Certainly there is no way for the computer to "look" at the pictures and make correlations.
-Terry
|
|
|
|
08/02/2004 06:56:04 PM · #4 |
There are so many similar images in everyday objects that it's hard not to have similar images next to each other.
|
|
|
|
08/02/2004 07:03:28 PM · #5 |
Originally posted by louddog:
There are so many similar images in everyday objects that it's hard not to have similar images next to each other. |
With many challenges, this is the case. Even if there are 10% of a particular "type" of image, say 30 out of 300, there is a very high probability that a few (that is, more than two) will end up next to each other in any given random series.
|
|
|
|
08/04/2004 01:02:11 PM · #6 |
I'm getting waaaay too many anomalies. Specifically I repeatedly get two of one item, and they are the only two in the challenge. This is rediculous. Whatever code your using isn't working very well. It is really hard to judge fairly when you have an identical subject to compare immediately after.
There is absolutely no way this is random.
|
|
|
|
08/04/2004 01:12:25 PM · #7 |
Random grouping for a particular voter is set the first time they view the thumbs, right? I've noticed that, if I go in, vote a few, then come back later, my "random" thumbs are still all in the same order as the first time.
|
|
|
|
08/04/2004 01:17:09 PM · #8 |
But if you skip a few, you bump the order around, so it will seem like you're seeing the same photos over and over. More of a problem as you winnow down the group.
|
|
|
|
08/04/2004 02:02:34 PM · #9 |
Alright, let me demonstrate. Assuming there are 300 photos, the odds of getting a given photo are 1:300. Assuming there are 30 flower photos. The odds of getting a flower photo are 1:10. The odds of getting a second flower shot immediately afterward is about 1:100.
Now, the odds of getting a given photo again are 1:300. Assuming there are exactly two of this subject, (let's say a photo of a cable modem - looked around my room to find something 'obscure' that isn't in a current challenge). The odds of getting the second cable modem immediately afterward is 1:89700. The problem is I should be seeing these anomalies once every 300 challenges or so. I see one in each and every challenge, if not more than one.
Steve - I only use the 'thumb' view once, and just vote straight through until I finish.
airatic - I don't skip any. I always vote before moving to the next image.
**edit** My figures might not be exact, but I think they are close, either way they give you an idea of the scale.
Message edited by author 2004-08-04 14:07:26.
|
|
|
|
08/04/2004 02:27:19 PM · #10 |
1 in a million chances happen every day...
Or to put it another way, statistically the chances of tossing a coin 100 times and getting 100 heads is exactly the same chance as getting 100 tails and also getting 50 heads and then 50 tails.
Humans have this knack of finding order in random occurances, and then take offense at the order. Consider that most people wouldn't pick lottery numbers 1,2,3,4,5,6,7 as they would be 'more unlikely' to come up than say 23, 51, 74, 1, 4, 56 & 99 (say out of the set 1 to 100) when in actual fact, either set is equally likely to show up.
Or to put it another way - how many times do you notice that you _dont_ get two similar photographs in a row ?
Message edited by author 2004-08-04 15:01:27.
|
|
|
|
08/04/2004 02:31:42 PM · #11 |
I'm fully aware of what your saying. However, it becomes less and less likely to be random each and every occurance. It defies the odds if you will.
To answer your question, it seems to me that whenever there are exactly two items of identical subjects in a challenge I get them in a row. If there are three, doesn't happen.
It's not that I'm getting two similar photographs in a row, it's that I'm getting the ONLY two of those photographs in a row, each and every challenge.
-----
Further:
For two identical items (where there are only two of them in the challenge) the odds of getting them in a row are 1:89,700. Having the same thing happen next challenge, the odds are 1:26,910,000. The next challenge 1:80,730,000,000. The next challenge 1:24,219,000,000,000.
Etc etc etc. I'm 100% positive this happens at least once per challenge.
All I'm saying is the random(code) needs fine-tuning.
Message edited by author 2004-08-04 14:46:59.
|
|
|
|
08/04/2004 02:57:14 PM · #12 |
Originally posted by jadin:
I'm fully aware of what your saying. However, it becomes less and less likely to be random each and every occurance. It defies the odds if you will. |
No it doesn't. Which is exactly what I am saying, and what you are missing.
Put it another way, consider how many potential pairs of images there are in any given challenge - then realise that as you vote on images, the odds of pairs of images appearing side by side actually increases, it is almost inevitable to see the results you are getting, either with a pseudo random (as it is in this case) or pure random selection approach.
Message edited by author 2004-08-04 15:03:49.
|
|
|
|
08/04/2004 03:02:26 PM · #13 |
The randomization code could not be "broken" and get the effect you are claiming. Doing it completely random would be a much easier code than making like subjects appear sequentially. You seem to be suggesting that the people who coded the site did it on purpose. They couldn't "accidentally" make that happen.
|
|
|
|
08/09/2004 01:47:15 AM · #14 |
getting old and annoying.
|
|
|
|
08/09/2004 02:40:53 AM · #15 |
Originally posted by jadin:
Alright, let me demonstrate. Assuming there are 300 photos, the odds of getting a given photo are 1:300. Assuming there are 30 flower photos. The odds of getting a flower photo are 1:10. The odds of getting a second flower shot immediately afterward is about 1:100. |
I thought this too untill I took my first statistics course in college. The odds on getting that second one in ten shot are..... one in ten! the third? one in ten. If you have flipped a coin one hundred times and they all came up heads, on the next turn the odds are? If you said fifty fifty you would be right.
Now the odds of getting one hundred and one heads in a row are astronomical, but once you have the first hundred behind you, they don't change the odds of the next flip one tiny tiny bit.
Odds are cumulative only when you are projecting forward in time, once you are half way through a sequence ( you have seen two macro bugs on a flower in a row ) the odds of getting another on the next random selection are not influenced at all by the previously voted image.
If you don't belive me I would like to invite you over some time to play blackjack or craps ;) |
|
|
|
08/09/2004 08:50:32 AM · #16 |
Originally posted by jadin:
getting old and annoying. |
Jadin: Be sure and read what StevePax wrote. He is 100% correct. If you are seeing groupings of similar shots, it is that way because of statistical probability, not for any other reason. |
|
|
|
08/09/2004 09:02:49 AM · #17 |
Originally posted by BrennanOB:
I thought this too untill I took my first statistics course in college. The odds on getting that second one in ten shot are..... one in ten! the third? one in ten. If you have flipped a coin one hundred times and they all came up heads, on the next turn the odds are? If you said fifty fifty you would be right. |
While what you are saying is true (that each image selection is independant), in this case it is actually ever more likely. The odds actually increase each time, because images are not replaced in the selection pool each time, so the remaining images are taken from a decreasing set. |
|
|
|
08/09/2004 09:47:36 AM · #18 |
The odds to me seem to decrease that you will pull similar subjects from a finite set.
If you have a challenge with 100 entries, 4 of which are flowers, the odds of pulling a flower on the first vote is 1:25. The odds of pulling another flower right after that, with the pool reduced to 3 flowers in 99 is 1:33. And again, after that the odds are 1:49, and finally 1:97.
So the odds are decreasing that you will pull subsequent flowers, because there are less of them in a large pool to begin with.
Even if you manage to avoid flowers until the set is smaller - say you manage to get through 50 votes without getting any of the 4 flowers (statistically unlikely but certainly possible). The odds of hitting 2 flowers in a row are still reduced, from 1:12.5 for the first flower to 1:16.3 for the second and so on.
Having said all that, of course I still think that all of the thumbnails are presented at random and that similarities just happen. :-)
|
|
|
|
08/09/2004 09:49:56 AM · #19 |
| Iam struggling to see what it matters anyway, no two images are the same and each should be taken on its own merit when voting. |
|
|
|
08/09/2004 09:51:28 AM · #20 |
Just to throw more fuel on an already ridiculous fire:
Most "random number generating scripts" aren't.
To build a truly random number generator is extremely difficult. Computers have an inherent problem with randomality since everything they do is programmed and to be random is to be "un-programmed". Most RNGs are a function of user input to "randomize" the number output. In this case, my guess would be that your user name is utilized somehow to randomize the order in which you see images. Some of the better RNGs use mouse position coordinates which means the more you move your mouse the more effective the randomization is.
Now I"m getting a warm fuzzy feeling thinking about all you paranoid delusionals sitting there wiggling your mouse complaining that you keep seeing similar images and how "Big Brother" is trying to influence your vote. Keep wiggling!
|
|
|
|
08/09/2004 05:39:14 PM · #21 |
Okay,
While the odds of getting a particular image are ever more likely. The odds of some particular event that has already happened is calculatable.
Say you have 10 numbered cubes in a box. The odds of pulling the number 1 are 1:10. The odds of pulling the number 2 are 1:10 etc etc. After the first cube is removed, the odds are now 1:9 of any particular number, save the first chosen number.
Now lets say you wound up pulling the 1,2,3,4,5,6,7,8,9, and 10, in this order on your first try are 1:3,628,800.
Now to get a given image the odds are 1:300, to get a given image after the first one, 1:299.
Those are FUTURE odds of getting something. I'm talking about odds of something that has ALREADY HAPPENED.
The reason why the odds of these events are so high, is because they've already happened. The 1:300 odds do not apply to this. I get "anomalies" each and every challenge. Therefore they are calculatable as things already occuring.
Let me demonstrate why the anomolies are more than just "oh two images of flowers in a row". My anomalies are more like the first image is titled "heads", and the next image is a picture of the tails side of a coin. They are related, not by a simple link, but by a unique attribute that becomes apparant only when next to each other. Let me give you another example. (I have real examples but cannot post them due to current challenges) First image is a duck, second is a duck, third is a goose. These are the types of anomalies I get. If it was as simple as three flower shots in a row, I'd say, oh well, it's bound to happen, but I get WEIRD anomalies.
My concern is that the random code as pfellner brings out, is that it's using something it shouldn't be to random(), such as random(title) or random(category). Please don't tell me what the server is using unless you actually programmed it, I can speculate just as easily, but it doesn't mean it's right.
Message edited by author 2004-08-09 17:40:09.
|
|
|
|
08/09/2004 05:49:28 PM · #22 |
Jadin what does it matter just vote on the image that you are viewing thats not hard to do.
You need to be able to do this, on other sites ive had to vote on subject matter that is quite specific, you need to be able to assess each image on its own merits. Even here in the portrait challenge and iam sure in other challengers the subject matter is narrow just learn to look at each one and discard what you have seen previous. |
|
|
|
08/09/2004 05:51:12 PM · #23 |
Originally posted by jadin:
My concern is that the random code as pfellner brings out, is that it's using something it shouldn't be to random(), such as random(title) or random(category). |
Do you really, honestyly think that Drew and Langdon would intentionally code the site such that images that were somehow related by title or category would be displayed next to each other? Come on!
The behavior you are seeing is completely random (within the limits of computerized random number generators). There is no "special ordering" of the images occuring because of title, category, or some far-fetched "image content similarity algorithm".
The images are displayed in a random order, courtesy of PHP's rand() function. Simple as that!! And note that every time you cast a vote and return to the thumbnail page, your "personal" photo order is re-calculated. (If you still have a challenge with a lot of non-voted entries, try it for yourself if you don't believe me...)
Message edited by author 2004-08-09 17:52:44. |
|
|
|
08/09/2004 06:27:04 PM · #24 |
Originally posted by jadin:
Let me demonstrate why the anomolies are more than just "oh two images of flowers in a row". My anomalies are more like the first image is titled "heads", and the next image is a picture of the tails side of a coin. They are related, not by a simple link, but by a unique attribute that becomes apparant only when next to each other. Let me give you another example. (I have real examples but cannot post them due to current challenges) First image is a duck, second is a duck, third is a goose. These are the types of anomalies I get. If it was as simple as three flower shots in a row, I'd say, oh well, it's bound to happen, but I get WEIRD anomalies.
|
I don't mean this to be insulting, but you do realize that you're the one seeing these patterns and clusters, not the server. The human brain has an amazing ability to find patterns - so much so that it often "sees" patterns when there are none present. If you go looking for weird connections, chances are you'll find them.
|
|
|
|
08/09/2004 06:29:42 PM · #25 |
Originally posted by EddyG:
The images are displayed in a random order, courtesy of PHP's rand() function. Simple as that!! And note that every time you cast a vote and return to the thumbnail page, your "personal" photo order is re-calculated. (If you still have a challenge with a lot of non-voted entries, try it for yourself if you don't believe me...) |
If, however, you leave the voting screen altogether (like go back to the home page), then click on the "Cast Votes" link again for that challenge, your original thumbnail order returns, except that any you have voted on disappear.
|
|
