Which of your colours stop being two colours

The number compared is the hue and chroma part of the difference — ΔE00 with the two lightnesses set equal — because that is the part a dichromacy removes, and the part colour-coded interfaces rely on. A pair is listed once it loses 80% of it, which is our threshold and not a published one; the exact percentage and the surviving lightness difference are printed for every pair, so you can apply your own.

    What the simulation actually does

    protanopia

    The long-wavelength (L) cone is missing.

    Confuses red against green, and red against black — red loses most of its brightness as well as its hue.

    deuteranopia

    The medium-wavelength (M) cone is missing.

    Confuses red against green, while brightness stays roughly where a trichromat expects it.

    tritanopia

    The short-wavelength (S) cone is missing.

    Confuses blue against green, and yellow against pink.

    Each of these is a projection: every colour collapses onto the plane the two remaining cones can still span. That is testable rather than artistic — white must come out white, and running the simulation twice must change nothing the second time. Both are asserted over the sRGB cube before a page is built.

    These are the three dichromacies, where a cone type is absent. The far commoner anomalous trichromacies — where the cone is present but its sensitivity is shifted — see more difference than this shows and less than you do, and no single picture represents them. So the page names the conditions it models and does not claim to show "what colour blind people see".

    The example the palette above starts with

    #d62828 and #2a9d3f are ΔE0068.9 apart and sit106° apart in hue — nobody would call them the same colour. Under deuteranopia their hues are 2°apart: the same hue, twice. What is left is ΔE00 6.3, and all of it is brightness. That is why a status light, a chart series or a form error cannot carry its meaning in hue alone.

    HexNormalprotanopiadeuteranopiatritanopia
    #d62828
    #2a9d3f
    #1d6fb8
    #f4a300
    #6a4c93