Color perception test scored by how far each chip sits from home

No — it measures discrimination and refuses to name a type, and the distinction is not modesty. A test that says deutan rather than protan is reading which of two nearly parallel confusion families your errors fall on, and that reading needs stimuli whose coordinates are known to land where the design says they do. On an unprofiled display in an unlit-or-lit room, the same chip can arrive anywhere within a visible margin of its specification, and the margin is wide enough to move errors between families. So the page reports the rows separately, names the pairs you transposed, and leaves the classification to somebody with a calibrated light box.

Because a row that varies in brightness can be ordered without seeing color at all, which is the commonest way a browser hue test measures nothing. Hold luminance constant and the only cue left is chromaticity. This is also why the three hue rows look washed out compared with the saturated wheels other sites use: a fully saturated red and a fully saturated blue are nowhere near the same luminance, and any row containing both is a lightness ramp wearing color.

Nothing, and that is its job. It is the control that separates two very different reasons for a high score: not seeing the hue difference, and seeing it perfectly well while finding the arrangement task itself fiddly. Somebody who scores 40 on the hue rows and 38 on the gray row has told you about ordering ten things, not about their cones. It is also the whole test for anyone who wants it — running the gray row alone is offered on the start screen as an equal choice, and the page says plainly that the question has changed to how fine a lightness step you can sort.

The first one, and the gap between them is learning rather than eyesight. An arrangement task is heavily practiced: on a second pass you remember roughly where the awkward chips went, and on a third you remember the row. The chips themselves never change between runs — only the starting shuffle does — so the useful comparison is not run one against run two but the same run repeated in a month, or your hue rows against your own gray row. The page keeps a seed with each result so an identical shuffle can be replayed deliberately.

It works, with two caveats that both make the rows easier rather than harder. Modern phones are wide-gamut and aggressively color-managed, which tends to render these sRGB chips accurately but at a higher peak brightness than a desktop panel, and higher luminance improves discrimination. The second is size: chips a centimeter across sit on more cone-dense retina than chips three centimeters across at arm's length, and small adjacent patches with a shared border are the easiest arrangement there is. Compare a phone score with a phone score.

Because it does, and simultaneous contrast is doing it. A patch surrounded by a slightly greener neighbor shifts perceptually toward pink, which means the arrangement you are building changes as you build it. This is a genuine property of the visual system rather than a defect in the test, and it is one reason the original cap tests keep the caps in a physical tray with a fixed gray surround. If a chip stops looking right after you move it, move it back and look at it against the two chips it will actually sit between.

Show them that you took it and what it did, not the number. Clinical color vision testing uses plates or caps of known spectral reflectance under a standard illuminant, usually a Macbeth easel lamp, precisely so that the stimulus is not a variable — and an anomaloscope, which is the instrument that actually classifies a deficiency, works by asking you to match a mixture of two monochromatic lights. None of that survives a browser tab. What a result here can honestly do is prompt the appointment.