Advertisement
Guest User

4chan Random-Picks-Grid FAQ

a guest
Oct 5th, 2019
2,634
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
XML 8.46 KB | None | 0 0
  1. 4chan Random-Picks-Grid FAQ
  2.  
  3. <!-- What is a random-picks-grid? -->
  4. A random-picks-grid (RPG) is a 3x3 of 9 items that have been picked from a pool of 3x3 automatically by a random number generator (RNG).
  5.  
  6. <!-- Which grids posted in a thread do qualify for selection? -->
  7. All favourite-anime-3x3 and all favourite-manga-3x3 can qualify, as can mixed grids with both anime and manga on them.
  8.  
  9. <!-- Which grids do not qualify and for what reason? -->
  10. Grids with more than 9 picks do not qualify, because the selection algortihm requires 9 possible picks on a grid.
  11. 3x3 that are not /a/ themed or contain non-/a/ picks do not qualify as they are not compliant with board rules.
  12. Character-3x3 do not qualify, unless the thread they are posted in is specifically stated to be a character-3x3 thread (unless the characters represent their respective anime/manga).
  13. Inverse-3x3, if they can be identified as such, do not qualify, unless the thread they are posted in is specifically stated to be an inverse-3x3 thread.
  14. Troll grids or meme grids, if they can be identified as such, do not qualify for obvious reasons.
  15.  
  16. <!-- If I post more than one anime-3x3 or more than one manga-3x3, will they all qualify. -->
  17. No, only one anime-grid and one manga-grid per poster. The one posted first will usually qualify over the one posted later.
  18.  
  19. <!-- Do images containing both an anime-3x3 and a manga-3x3 qualify. -->
  20. Yes. The anime-grid and the manga-grid will be entered into the selection pool seperately.
  21. If there are more than one anime-grid or more than one manga-grid, the upper-left most will be selected by default.
  22.  
  23. <!-- How can you tell whether something is a character-grid, an inverse-grid or by the same poster as another grid. -->
  24. I cannot. Not reliably anyway. Therefore I will question posters whenever I'm in doubt and try to only ever disqualify grids in obvious cases.
  25.  
  26. <!-- How can you pick 9 grids randomly with a RNG? -->
  27. Initally, once there is a certain number of grids in a thread, they can be arranged in a list.
  28. This list can be ordered randomly by a "random shuffle".
  29. More items can then be added to that list in a random position by a "random insertion".
  30. Each item on the list also gets a random number from 1 to 9 denoting one specific panel to be selected from that grid (according to their positions on a phone numpad, with 1 at athe top left).
  31. At all times, the first nine items on the randomly ordered list will constitute the RPG.
  32.  
  33. <!-- What is a "random shuffle"? -->
  34. In a random shuffle, for each postition i of a zero-indexed array of the length n, a random whole number j is generated in an interval [0,n-1] and the element at i is swapped with the element at j.
  35. In other words: Each element gets swapped to a random position at least once. This results in a randomly ordered list where each element has the same propability to be at any position.
  36.  
  37. <!-- What is a "random insertion"? -->
  38. A random insertion inserts an (n+1)th element into an array of length n in a random position. For that purpose, a random whole number j in an interval [0,n] is generated and used as the position for the new element to be inserted at.
  39. Insertion itself can either be done by shifting all elements from that position j onwards by 1 and writing the new element into the now "empty" spot at j. Or inserting the new element at position n (the end of the array) and swapping it with the element at j.
  40. In our case, the second option, swapping, is used, as it only affects one other element's position at most, thus leaving most of the grid untouched.
  41. Both approaches guarantee that the uniformly random distribution of the array's elements is kept: If each element in the input array had an equal chance of 1/n of being in each spot, in the output array each element will have and equal chance of 1/(n+1) of being in each spot.
  42.  
  43. <!-- What kind of RNG is being used and is it really random? -->
  44. The RNG used is .NET's default RNG, which uses an implementation of a subtractive RNG algorithm.
  45. Like all deterministic algorithms, it is not truly random, but only pseudo-random, which means it creates a sequence of numbers that appear largely random, based on a seed that is created from system variables at the time of the program's startup.
  46.  
  47. <!-- Why does a grid only ever chance in one or two places when an update is posted? -->
  48. As each update inserts only a few new grids at a time using the "random insertion" algorithm, which is designed to affect the grid as little as possible while still guaranteeing a uniformly random distribution, the grid will rarely ever chance in more than one or two places at once, if at all.
  49. Furthermore, the more grids in the selection pool, the lower the chance that a newly added grid will get selected: The 27th grid of a thread has a 33.33% chance of being selected, whereas the 36th grid only has a 25% chance. (They do, however, at each given point in the process, both have the same chance of having been selected or not selected.)
  50.  
  51. <!-- How do you read the "raw data" numbers that accompany an RPG? -->
  52. Source grids are identified by their post-ID. That's also how they appear in the raw data, sometimes with a further identifier like an "a" or an "m" to distinguish different grids that were posted together in one picture.
  53. Each ID will have some +(n) next to it, where n is a number from 1 to 9 that denotes which panel was selected.
  54. Raw data given is limited to the first nine items on the random list, as only those 9 are relevant for the RPG.
  55.  
  56. <!-- What is the minimum number of 3x3 in a thread for a RPG to be made? -->
  57. The threshold for each thread is usually 24 eligible grids being posted.
  58. If that threshold has been missed two times in a row, the next thread will have a lowered threshold of 22 grids.
  59. If a threshold has been missed three times in a row, the next thread will have a threshold of 20 grids.
  60. If a threshold has been missed at least four times in a row, the next thread will have a threshold of 18 grids.
  61. The threshold will never go lower than 18 (50% chance for each 3x3 to be picked).
  62.  
  63. <!-- What happens if the same anime or manga is picked from two different grids? -->
  64. In the rare case that the same thing is picked from two grids (and I notice that), I will randomly select another pick from the grid with the dublicate by by randomizing yet another number from 1 to 9 for it.
  65. Should I miss a dublicate, I do appreaciate being made aware of that error.
  66.  
  67. <!-- Why do certain anime/manga show up more frequently than others? -->
  68. Certain picks are more common than others. A popular anime that appears on five of 27 grids in a thread has a 18.5% chance of being selected, but a rare pick that's only on one of 27 grids has a mere chance of 3.7% .
  69. That's because  1 - (1 - 1/9 * 9/23)(1 - 1/9 * 9/24)(1 - 1/9 * 9/25)(1 - 1/9 * 9/26)(1 - 1/9 * 9/27) = 1 - (22/23)(23/24)(24/25)(25/26)(26/27) = 5/27 = 18.5185%
  70. and  1/9 * 9/27 = 1/27 = 3.7037% .
  71.  
  72. <!-- Why are certain grids selected more often than others? -->
  73. Because certain grids get posted more often than others. A grid that gets posted in practically every 3x3 thread will have a higher chance of being selected at one or more points (over the course of a longer timespan) than a grid that only gets posted once a month.
  74.  
  75. <!-- Speaking of chances, what are the chances of my grid being picked? -->
  76. At any given moment, if there are n grids in a thread (for n >= 9), your chances are 9/n and 1/n for each specific pick on your grid.
  77.  
  78. <!-- But I have both a anime and a manga grid. What are my chances then? -->
  79. If you have two eligible grids, the chance that one of them gets picked is (1 - (1 - 9/(n-1))*(1 - 9/n)) = (18n - 90)/(n^2-n) , which is still over 50% for n = 30 .
  80. The chance that one specific pick of yours gets picked is (1 - (1 - 1/9 * 9/(n-1))*(1 - 1/9 * 9/n)) = 2/n
  81. and the chance that both your grids get picked is (9/(n-1) * 8/n) = 72/(n^2-n) , which is around 10% for n = 27 .
  82. However, the chance for two specific picks of yours being selected is merely ((1/9 * 9/(n-1)) * (1/9 * 8/n)) = 8/(9(n^2-n)) , which is around 0.1% or a 1 in a 1000 chance for n = 30.
  83.  
  84. <!-- Where can I find older RPG? -->
  85. There is a MEGA archive with all older grids here:
  86. https://mega.nz/#F!VvAlzYzT!32Ca0qvfqMjMYsGu9yUn6g
  87.  
  88. <!-- Why are you even doing this RPG thing? For what purpose? -->
  89. For fun, mostly. But also because it's interesting to get random samples of /a/'s taste.
  90. Despite the small sample size (the bigger the selection pool) and a lot of the grids being from the same "regulars", I've found the RPG quite varied and representative so far.
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement