CS Rarity List [WIP] - status update on page 75

[Erythro]

Sorry guys, I honestly don't have much time to run the script (which will include remembering how it works and how to run it OTL. I recall that a big part of it was manually saving transparent images of each pet.)

If someone else is willing to take up the work, it'd be easiest to rewrite the script. If you need help with image processing or something, I can write a description of the way I did it.

If transparency is something that needs to be done, there is a website here with most CS transparencies: https://cstranspets.weebly.com/
I know very little about coding. (Literally only know how to make a matrix in R (stats program)) However, if someone else knows coding like that, I think it would be an absolute huge help.

[nickjr]

Law, you used Python, right? I know a little bit of Python. I'd need instructions on how to build and run the script though

[Law]

It's more complicated than just running the script, unfortunately.... :(

[nickjr]

I can probably follow the instructions but I'm not sure

[Davy + Aero]

Law - out of curiousity, how did you do it? I could try to throw together a script in python if that's how you did it (I KNEW I WOULD BE SOMEWHAT USEFUL) but how?

--Aero

[Law]

Law - out of curiousity, how did you do it? I could try to throw together a script in python if that's how you did it (I KNEW I WOULD BE SOMEWHAT USEFUL) but how?

--Aero

Basically I have every pet image saved:


Each image is preprocessed by cutting off the part that says "_'S PET" (cropping the bottom pixels), and resizing to a smaller size. These are the reference images.
The reference list contains [pet's image, pet's original date, pet's name] for each pet.

~~

To get the pet counts, I randomly choose PET_IDs, and get the images from: http://www.chickensmoothie.com/pet/{PET_ID}&trans=1.jpg

These images are preprocessed in the same way (cropped, resized) as the reference images, and are compared to each reference image (there's a smarter way of comparing, but this is generally what happens). Once the matching reference image is found, that specific pet's counter is incremented.

[Solloby]

On mobile atm but thanks for the explanation Law. Do you also include &notiems or W/E it is so dressed up pets count?
Side note, in future we will need to check pet page for onsite/offsite I think. Many people adopt offsite rereleased, and this may over exaggerate the impact of rereleases on the data.

[Law]

On mobile atm but thanks for the explanation Law. Do you also include &notiems or W/E it is so dressed up pets count?
Side note, in future we will need to check pet page for onsite/offsite I think. Many people adopt offsite rereleased, and this may over exaggerate the impact of rereleases on the data.

Oh yeah, that's right. I used "&noitems".

[Lya]

edit: Has anyone heard from Lya recently?

whispers: I'm here, lmao. Needed me for something specific? ^^

I'm currently a bit sleep deprived, so I can't really comprehend the more complicated stuff that's going on, but I'll read through the topic in the next few days.

Anyway, with what I can comprehend, I'm amazed that you're working with an extra program for this. (I'd be that person trying to run automatical counting through a google spreadsheet - only data putten in by hand would be ID range, if the pet is relevant and for the relevant pets the name. And all this without having to leave the sheet even once.)

Edit:
I'm a bit more awake now and have read through the first few pages. Saw that you tried a spreadsheet. (Though you could have just made it display the pet images instead of having to open each pet in a seperate link?)

Edit 2 (finished 5 hours after post time):
Okay, I finished reading the whole thread. @_@
A few thoughts from me - please excuse if anybody said the same thing before. I'm also "simplyfying" a few statements I saw that weren't clear to everyone.

For analysing the data we already have:

• Tess confirmed that the chance to get the new URs is higher than the chance to get an old UR.
  
  Let us say there is a sample of 1000 pets adopted during a rerelease and the chance to get an old UR is set to 0.1%. This means out of these 1000 pets, one pet should be this UR.
  Now, a new UR has a higher chance to be adopted. Let's say this chance is set to 0.2%. This means out of these 1000 pets, two pets should be this new UR.
  
  So much for adoption rates within rereleases:
  +1 old UR
  +2 new URs
  
  However, this does not mean that there are twice as many of the new URs on the site than of the old UR.
  
  What truly happens with rarity:
  
  Let's say the older UR was released the previous year with a 0.2% chance of getting adopted, too, and that there, too, were adopted 1000 pets.
  
  So the number of old URs before the rerelease was 1.
  After the rerelease, the count for the old UR is 2. The count for the new UR is 2, too.
  
  This means, after the rerelease, both URs would be equally rare, because there is an even number of them.
  
  Why I'm explaining this:
  I would like to believe that the rerelease and UR release chances are tweaked in a way to produce the originally planned rarity again.
  
  In an example that makes this clearer:
  
  There are 100 000 pets total. Out of these 100 000 pets, 100 pets are Sunjewels.
  This makes for a rarity of 1 in 1000 (0.1%).
  Tess originally intended for the Sunjewel to have a rarity of 2 in 1000 (0.2%).
  To accomplish this, she needs to make sure a certain number of Sunjuwels gets released in the rerelease.
  
  If 100 000 pets get released in the rerelease, we are at 200 000 pets total.
  If the Sunjewel is released like it was originally intented at 2 in 1000, there will be 2000 new Sunjewels released.
  We would then be at a total of 3000 Sunjewels in 200 000 pets.
  This would be a rarity of 1.5 in 1000 (0.15%) and therefore not what Tess wanted.
  
  She has to tweak the release chance!
  
  If there are 200 000 pets after the release and 0.2% of them are meant to be Sunjewels, there need to be 4 000 Sunjewels after the rerelease.
  Since we already have 1 000 Sunjewels, there need to be 3 000 new Sunjewels released.
  3 000 Sunjewels out of 100 000 pets that get newly released means that the chance to get a Sunjewel has to be set to 0.3%.
  
  And that's what I think is happening (more or less).
• It's important to keep in mind that the above only works for rarities before 2012, because it relies on the total number of pets only.
  All rarities after 2012 rely on total number of pets on active accounts which is a variable we have no means to identify.
  However, we can make assumptions based on the percentages within each rerelease.

And afafejlafinlavbiles.... I managed to totally jumble up two different thesis' in there, I'm sorry.
I might get this into actual usable form later. *facepalm*

[nickjr]

YOU'RE ALIVE HI