Thank you, do you mean it is darker than it should be instead of lighter? Because the top is the second lighter shade. The reason I made it darker is because the light is supposed to come from an upper angle, slightly higher than 45º from the ground, so that the right ramp facing up is the brightest, but still not getting super light from glare.

The main error seems to come from the fact that I tried to reuse colors, when it should actually be one color for each specific angle of the faces. In any case there seems to be 9 visible angles, but 9 shades of every color ramp from dark to mid tones is way over what I am comfortable to work with. Even if I connect the color ramps and reuse the last two darks it would still demand a larger palette than I would like.That is why I tried to reuse the shades on faces that wouldn't touch each other, but as you can see, that is difficult to predict, there can always be a construction that would make them touch and break the sense of volume.

For instance, I know the top right ramp on the larger "dome" should be a lot darker considering the angle of light, but whenever I painted it darker, it didn't looked like a opaque prism anymore, and was implying some sort of crystal. The makeshift lighter tone seems to be acting like a back light or difuse light somehow, and adds to the shape, but it is incorrect and wasn't intentional.

Thanks again, I think I need references to study since instinct alone isn't making it.

Sometime ago Cyangmou gave me a great lesson on color palettes and color in general when I tried to calculate a palette keeping values constant across all hues and failed bad. Since then I have been on and off about trying to make a more personal palette, learned to love and respect saturated reds and yellows but remained still interested in the tight swatches of supergameboy's/gameboy color 4 shades mixing and the similar results I was getting with a few colors from nes ntsc palettes. I finally tried to nail down what I was really trying to achieve, and realized I wanted a tight, four shades at most, somewhat sepia palette, that would give me a pokemon on gameboy vibe but with higher hue variation and lowered saturation, resembling colored pastels over a drab, darker paper. I decided I would be strict in only keeping useful colors, something I learned here. This is what I came up, 11 colors in total:

I am still uncertain about the yellow ramp, and in doubt if it deserves its own third shade, since I am using a greyish dark purple instead of brow or dark olive. I think one of the keypoints in making this colors work is use a lot of the brightest color in the sprites, and the contrast of the drab "white" and the second shade of yellow is not ideal yet. I value you guys opinions a lot, and would like any input you might have on anything that might improve this. Thanks in advance.

Thank you very much Cyangmou for the lengthy, insightful answer. Rest assured that, although a beginner with pixels, I can understand what you said from my background, and after reading it I do feel quite dumb for trying to come up with a more "rational" way to develop a palette, as if there was a shortcut to make it my own.

Somehow I assumed there was some secret about keeping value constant that made palettes look more balanced, and I was going for something like that here, in a very failed experiment, thinking it was somehow better than other palettes I developed more naturally. The brights indeed are more like placeholders than individual colors, they all look the same, almost the same for the darks. That was already the case with NES NTSC I started from, but looking at a lot of numbers I thought I had fixed them without actually reality checking, I don't know how I managed to do that, really sorry 

As Cyangmou explained very clearly, each hue have its own "quirks", and they just cant be treated the same, yellow is so inherently brighter to point it is no longer yellow if darkened, blue azure cant be made brighter without looking more like lilac and so on. Trying to handle yellow, purple and deep blue the same way I would handle "easier" colors in HSL have always been puzzling, and after a while I started to learn to compensate the differences with brightness, saturation etc.

I do have a basic understanding of that, and my previous palettes are probably more useful than this one. After reading a lot of "tutorials" I did became in love with the uniformity of color models that compensate the inherent brightness of each hue, and I thought I had something there

Here is a previous experiment on generating a palette where all colors would decay towards the same dark purple and rise towards the same light yellow, this time with a ramp generator (shame), I can now see that same mistake happens here:

In my defense, when I used it I restricted my self to half of the shades and used the others as intermediaries for AA. But there are a lot of very similar colors.

From now on I will refrain from trying to model palettes, I will review what I think I know about color, and I will try craft my palettes according to what I like, what works, and what I need.

Thanks, that is what I thought. The solution for properly pixelated light effects surely must be setting everything for the 1:1 internal resolution. The paletted lights and shadows, however is something I've never seen in action aside from 16-bit era videogames, and I am not sure it is worth the trouble (but it is sweet in link to the past shadowy dungeons, I refuse to believe they doubled the tileset and sprites just for shading). Having a palette swap system working, however, seems very useful for various reasons besides light effects. Oh, and sorry for hijacking the thread, I was just curious and I learn something everyday on these boards.

Hey I am back after a flu, here is the latest attempt, the more I look at real trees and references, the more I see repeated patterns, same-shape forms with regularity broken only by fusion or separation, so I tried to fuse more and got rid of the very small indefinite tufts, but it did loose a lot of texture. Maybe now it seems less tridimentional than the second one, but from the first attempt it is improved, and it there is much more darkened gaps implying branches. Thank you so much Meat for all the help.
first: last: