Finally, after far too long, I have finished writing a full breakdown of the random number generator in Red, Blue and Yellow, and how it affects capturing in R/B/Y. I think it's pretty fascinating! Please take a look and tell me if you can't follow the explanation at all. (But one of the primary takeaways is: those annoying Pokémon in the Safari Zone that seem near-impossible to catch genuinely are way harder to catch than intended.)
The R/B/Y capture calculator has also been updated accordingly (a little while back) and is a fair bit more elaborate and sophisticated, thanks to the relatively intensive calculations involved in simulating every possible RNG outcome; please let me know if you have any issues with the new one, before I convert the other capture calculators to this new system.
I actually haven't yet updated the Safari Zone calculator; that's coming too, I promise, but I just needed to get this up first.
My own messages will be signed as Butterfree, with the Admin label below my name. If someone signs as Butterfree without that label, it's probably not me.
The graphs chart a couple of particular special cases, with each individual data point precalculated by the calculator (as in, I literally used a calculator that works like the actual calculator to get the result for 100/100 HP, then 96/100 HP, then 92/100 HP… and then took that list and hardcoded it onto the page so I could tell Google Charts to plot it out). There is no simple predictive formula behind those graphs; these are just the numbers that you get when you plug these particular parameters into the calculator. If I wanted to make charts for any other catch rate, or for any situation where there's also a status condition, I'd have to calculate each data point there individually from scratch using the calculator.
The limit of 255 doesn't really have anything to do with making it difficult to generate true randomness. All random number generators used in games, etc. are pseudo-random - they consist of applying some kind of formula that gives results that appear random. This RNG design in particular just isn't fully thought-out; there are many other random number generator algorithms that would not have this problem and would genuinely give the expected results.
Commenting on: 12-09-20
REALLY impressive work from all of you who've looked into this! I would've never expected a limit of 255 to make generating true randomness so difficult! Or did they make it more complicated than it needed to be..?
Just an offhand wonder about the new calculator: you say it's a heavy machine now, but you already have stats onhand for the predicted theoretical & real performance of these balls caused by this rng system, that you even put into graph form. Is it that these stats weren't satisfactory to apply as plain percentages during the simulation.. or that they weren't satisfactory for your ambition? 😛
Haha, they weren't exactly made to be standalone! Like, I've been posting them in places as cool RNG art, because they just look cool, but I absolutely don't expect them to convey much of any information without the accompanying explanation. That's why they are presented with all that explanation!
I'm not sure that they'd be all that useful standalone even with labels - like, even if I labeled the different colors, you wouldn't actually know what the patterns mean and they would just make it harder to tell how much green there actually is, proportionally, right? I could also label the axes, but the axes also aren't going to mean anything to anyone who isn't familiar with RNG internals as explained in the article. At best it might make the image more intriguing, I suppose, as a thing to post alongside a link to the article?
Yooooo, congrats on finishing this! It came out really nicely, and I think you did a good job breaking down what's going on with the RNG and how it has an impact on capturing. The capture plots are a really cool way to visualize what's going on… you can definitely see when the RNG is having a big effect based on how chaotic they get, heh. They're fun to look at as little pieces of art!
It might be nice to put some labels/legends on them, though, since without the paragraph of explanation about what's going on they're basically impenetrable. Compromises the arty look a bit, but would make them more standalone, I think.
Commenting on: 12-09-20
I haven’t read through this yet but this looks amazing!! Nice work, this must have been quite the research/testing task :D
Page last modified November 2 2020 at 21:21 GMT