Chocobo dyeing R&D Thread - Post your results here!

[Giantbane]

The guts of the problem is the calculator makes its calculations based on each fruit feed, not the kind of food feed.

I don't know what you mean by this

What you have written which I deleted supports the 2nd case, but not the first.

everything I wrote can apply to only feeding berries or only feeding apples, or feeding both. I only used berries as an example.

As of now, we can all safely say when you feed deepen red, expect it to behave to make the bird more red, and when I feed deepen blue to that red bird, we surely can expect it to become purple.

well, you'll forgive me, but that's not particularly helpful.

What the calculator does is it leads people to believe that each unit of fruit feed has an expected value in the of the RGB scale. But the fact is we know we can feed only 1 apple and it may cause it to move from scale 1 to 2, and 20 apples to move from scale 4 to 5, while both 1-2 and 4-5 only differs by 16. So, what is the R value per unit of fruit in this case? 16, or 0.8? Now factor in the RNG bomb SE throws in, what is the exact behavior of each fruit fed? So, when we recommend someone to feed 5 apples, should we be expecting a +90R, or +3R? Then don't forget to add in the RNG (unknown).

I myself tried to duplicate the Royal Blue feeding from the old spreadsheet of 30 berries and 10 apples, and I only end up using 26 berries and 1 apple. Using my own case, should I consider each apple a +5R (10 recommended), or +50 (1 actual).

The problem, if they do use an internal RGB color, is that RGB 0-255 represents over 16 million different color values and they're crammed into only 85 different dye colors. Each dye would represent, on average, ~200 thousand different color values. So the recipes, even if there weren't any RNG, could vary wildly as you'd only need to hit the range of color that represents a dye. Throw in an RNG component and it gets more complicated.

Because of this, it might be difficult to pin down an exact range for the RNG. However, whatever the range is, it's got to average out to something. +5/-5 doesn't seem like a bad place to start. It well for me, but maybe I got lucky. If people use this idea properly, we might be able to get a better idea whether this consistently overshoots or undershoots the targets. There will be outliers, but you should hopefully be able to ballpark it.

As said earlier, my best suggestion ATM, is keep making recommendations, but instead of using the calculator, use the PNG file of known color paths. In addition, instead of recommend tons of fruits to feed at once, lets feed one kind of fruit at a time, so we can start to obtain more actually color changes and their stages based on one kind of fruit. Once we reach within the desired color range, then make the necessary fine adjustments.

The problem I see with the PNG file is that it's extremely imprecise. It doesn't take into account the RGB values at all. It tries to deal with the colors as states with clear pathways. But the pathways can be anything but clear in some cases. It works well when you're on a single fruit color path or very close to it (for instance getting to blood red is easy, just feed apples). If you deviate from it greatly, such as the case for Soot Black, then it gets harder to make it work. When you get into the darker colors, if you strictly go by feeding one fruit at a time, you're very likely to end up in some weird loops with only guesswork to get out of it.when it's constructed of very precise color paths.

I think careful use of the calculator is probably going to give better results than strictly going off the PNG file because it deals with how the colors relate to each other. But there is a problem I've seen when people use the calculator. They rely too heavily on the calculator's color picker, but because there is some RNG involved and it's very unlikely that the hastily crafted calculator's color picker matches the actual game algorithm. It's not enough to simply get the color picker to guess at the right color. You need to try to target the colors as closely as possible. And more than just getting the values numerically close you also want to try to keep the difference from the target value as small as possible. For instance, R + 6, G+0, B+0 is not as good as R+2,G+2,B+2 because it preserves the ratio of the colors to each other better.

The idea is that every exact dye color represents the center of a color space represented by that color. And you want to adjust where you are to as close to your target's center as possible. Due to the RNG and unknown exact values you are not likely to hit your target space every time, especially if you're far off the center of your own color space. But as you keep heading toward that center, the RNG will average out and you're likely to get closer every time until you hit it.

[AttacKat]

careful use of the calculator

And how do you propose to achieve that, when we have results like this?

Meadow Green (143,159,103) -> 4 apples, 2 plums, 12 berries (-10:-14:-10) -> Raptor Blue (95, 135, 199)

RGB delta = +48, +24, -96

How does that fit into the -10:-14:-10 (-5:-7:-5) ratio using the calculator when the actual result is 2:1:-4?

[Giantbane]

results like this?

Meadow Green (143,159,103) -> 4 apples, 2 plums, 12 berries (-10:-14:-10) -> Raptor Blue (95, 135, 199)

RGB delta = +48, +24, -96

How does that fit into the -10:-14:-10 (-5:-7:-5) ratio using the calculator when the actual result is 2:1:-4?

Meadow Green, (143, 159, 103)

add 4 apples, 2 plums, 12 berries to get

93, 89, 153

and have the color picker decide that (93, 89, 153) = Raptor Blue

not to mention that you didn't start at 143, 159, 103. You started at one of the some two hundred thousand colors within the meadow green color space.

And how do you propose to achieve that?

Simple, you try to make the "To Target" numbers as small as possible in order to aim for the "center" of the target color space. If you're aiming for the center of the color space, it gives you the most room for error caused by RNG and an uncertain starting point. "Current Guess" should be ignored entirely.

Advanced users can then further adjust by guessing where they are in their current color space based on what colors they've been to. But this is a little more complicated.

[AttacKat]

Simple, you try to make the "To Target" numbers as small as possible in order to aim for the "center" of the target color space. If you're aiming for the center of the color space, it gives you the most room for error caused by RNG and an uncertain starting point. "Current Guess" should be ignored entirely.

So, can you put that down in RGB values for ALL the 85 colors we have, so for each color, we can have an actual "target" to aim for, and a known "center" to look at. It will be a very helpful reference for others.

[Giantbane]

So, can you put that down in RGB values for ALL the 85 colors we have, so for each color, we can have an actual "target" to aim for, and a known "center" to look at. It will be a very helpful reference for others.

It's the list of RGB values for the dyes. Even if they're guesses and not exact, they still represent a pretty good estimate of how the RGB values of the various color spaces relate to each other. Which is really the most important thing, having a numerical representation for how colors differ and how fruits change that color. It allows you to, for instance, have a better idea of how a gloom purple differs from a regal purple. And if you want to move from one to the other, you can make an educated guess as to how you'll need to move. This seems better to me than just taking single steps. I'm almost positive that these color spaces do not exist on a clean RGB grid that you can just transition easily from one dye to another in an orderly fashion. Crossing the boundary between two colors may require you to move in multiple directions at once, and not necessarily at 1:1:1 ratios between primary colors.