Mixing two pigments subtractively removes more of the light spectrum than either pigment alone, which is why mixed paint colours usually end up less saturated ('muddier') than either starting colour. The exact result also depends on the specific pigments used — two different blues and two different yellows do not have identical reflectance, so 'blue plus yellow' does not give one single, fixed green.
Practical advice
Small test mixes on the actual materials involved are more reliable than assuming a theoretical mixing result, especially when combining paints from different product ranges.
| Model | Result | Lightness | Chroma | Hue |
|---|---|---|---|---|
| Cobalt blue (input) | #0047AB | 42.8 | 17.2 | 260° |
| Lemon yellow (input) | #FFF44F | 94.8 | 17.7 | 106° |
| Added as light, half and half | #BCB987 | 77.6 | 6.7 | 105° |
| Average of the stored sRGB values | #809E7D | 66.7 | 5.8 | 143° |
| Ideal filters in series | #004432 | 34.3 | 6.9 | 168° |
| Midpoint in OKLab | #81A39B | 68.8 | 3.9 | 179° |
Four defensible calculations give four different colours, from lightness 34.3 (ideal filters in series) to 77.6 (added as light, half and half), reading as yellow, green, blue-green. Every result has less chroma than either input — 3.9 to 6.9 against 17.2 and 17.7 — which is the loss of saturation mixing is known for. None of them should be taken for what two paints would make.
How this was calculated. Lightness, chroma and hue are OKLCH (lightness 0–100, chroma ×100). None of the four models predicts paint: a pigment mixture depends on how each pigment absorbs and scatters light at every wavelength, which an RGB value does not record. Predicting it needs measured reflectance spectra and a model such as Kubelka–Munk.