I used to simply have at it and see what happened, but in recent years I have gotten a bit more methodical. Inspired by such works as Richard Schmid's Alla Prima and James Gurney's Color and Light I would try to make neat and orderly grids of carefully calibrated tints and shades. I found it tedious and difficult to get the colors to stay inside perfect squares, so it wasn't long before I let the production values slack. I still try to put the colors through their paces, seeing how they lighten in a wash or darken when mixed with complements or near complements.
The 12 colors that came with the Schminke box, mixed with more and more water:
Burnt Sienna, a candidate for the tiny watercolor box, mixed with each of the other candidate colors:
Painting a color wheel is a great exercise to get the hang of new paints. I was content with a fairly slapdash product:
Recently, having learned some more about color theory and color spaces, I began looking for a way to get more flexibility and precision. Here the first iteration of my brainstorm:
The colors are the ones I eventually decided to include in my new Schminke watercolor box. Some were new Schminkes, some old favorites, and some rediscovered because they seemed to go with the Schminkes.
The design is based on Gurney's simplified color wheel with pigments which was in turned based on MacEvoy's more complex one.
I traced out the circles with a compass. The radiating lines divide the circumference into 30˚sements, like a clock face, and mark off the theoretical hue positions of Yellow at 12 o'clock, Magenta at 4:00 and Cyan at 8:00, for the subtractive primaries, and Red, Blue and Green, the additive or light mixing colors at 2:00, 6:00 and 10:00, with intermediate gradations in between.
The concentric circles represent levels of saturation or intensity, so colors closer to the center are increasingly grey or brownish.
Instead of painting the pigments directly onto the wheel, I brushed them out onto watercolor paper and cut the paper into 1/2" squares to make color chips. I then printed out the color names and made labels, and affixed each chip and label to the wheel with a pin.
The idea is to see how your actual pigments map onto a theoretical color space. I think my two examples contained elements of the CIELab and YRMBY hue circles and the Munsell scale of intensity. For my purposes, any consistent and easy to interpret system was good.
I started by following the layout in Gurney's and MacEvoy's examples, and tried to fine tune the positions for my particular pigments. It was interesting to note, for instance, that Quin Magenta, which I think of as my "true" magenta for mixing purposes, is actually in between true magenta and red on the wheel. It's also interesting to see how yellows and reds have more degrees of saturation and max out near the circumference at bright values, while blues and violets max out near the center and dark.
I'm just getting into this, but is seems like a useful exercise.