Three difficulty tiers
Easy uses bold, obvious hues. Medium uses named tones that sit close together — teal, turquoise, indigo, coral. Hard uses designer shades like periwinkle, puce and verdigris, where knowing the name is half the battle.
One color fills the screen and four names sit under it. Pick the right one. What makes this a color vision test rather than a vocabulary test is where the wrong answers come from: some are chosen to be colors that collapse onto the target only through a specific deficiency — protan (red-weak), deutan (green-weak) or tritan (blue-yellow). Reaching for one of those is direct evidence, not a near miss.
This is the closest thing on the site to how an Ishihara plate works, and it is built on the same idea: show something that looks like one thing to typical color vision and something else entirely to a particular kind of colorblind eye. The difference is that a plate is a fixed image with a fixed answer that plenty of people have already memorised, while every round here is generated fresh. The wrong answers rotate through all three confusion axes over a run, so the evidence stays balanced instead of hammering one of them and going quiet on the other two.
Easy uses bold, obvious hues. Medium uses named tones that sit close together — teal, turquoise, indigo, coral. Hard uses designer shades like periwinkle, puce and verdigris, where knowing the name is half the battle.
Rounds run about five seconds each. Ten is a quick game and supports only a low-confidence reading; fifty takes five minutes or so and is where the color-vision estimate is at its strongest.
Not a diagnosis. Your screen is uncalibrated and your room lighting is uncontrolled, and the quiz cannot separate "did not recognise the color" from "does not know the word". A result here is a reason to see an optometrist, never a replacement for one.
Prefer mixing colors to naming them? Chroma Match gives you red, green and blue sliders and a target to rebuild. It carries a richer signal per round because your answer is a point in color space rather than one of four choices, and it never depends on knowing what "verdigris" means.
Both games are free, need no account, and run entirely in your browser — nothing you do here is uploaded anywhere. Pick whichever suits you: one measures how precisely you can reproduce a color, the other how reliably you can name one.
A target color appears and you rebuild it with red, green and blue sliders. Because your answer is a point in color space rather than a choice from a list, every round records exactly which direction you drifted in — which is what the color-vision estimate reads.
One color, four names. The wrong answers are not random: some are picked to look nearly identical to the target only through a specific color-vision deficiency. Choosing one is a much sharper signal than a near miss on a slider.
Play it straight for a meaningful result. Take each round seriously, turn color assist off, and use as many rounds as you have patience for. Rushing or clicking at random is detected and reported as an unreliable run rather than dressed up as a finding.
The short version: sloppiness scatters your errors evenly in all directions, but a color-vision deficiency stretches them along one particular axis. The test measures that stretch.
When you finish a round we keep the signed difference between the color you produced or picked and the color you were shown. Over a run those differences form a cloud of points in color space.
Ordinary human imprecision is roughly the same in every direction, so the cloud comes out spherical. If two colors look identical to you, you can slide freely between them without noticing — so the cloud stretches into a cigar shape pointing along that confusion axis.
For each of the three confusion axes we divide the spread along the axis by the spread in the plane across it. A perfectly even cloud scores 1.0 no matter how sloppy or precise you are — which is what makes the measure fair to careful and careless players alike.
That ratio becomes a z-score whose reliability depends on how many rounds you played. Because we compare three axes and report the strongest, the threshold is raised to compensate — otherwise one of the three would look meaningful roughly one run in seven purely by chance.
This is not a cosmetic label. The precision of the measurement scales with the number of rounds, so a longer run can support a stronger claim. Simulating 400 players of each kind at each length gives the following:
| Player | 10 rounds | 20 rounds | 30 rounds | 50 rounds |
|---|---|---|---|---|
| Typical vision, wrongly flagged | 4% | 8% | 7% | 4% |
| Deutan correctly detected | 91% | 97% | 99% | 99% |
| Protan correctly detected | 88% | 97% | 98% | 100% |
| Tritan correctly detected | 89% | 97% | 99% | 100% |
| Random clicking | unreliable | unreliable | unreliable | unreliable |
| Confidence reported | low | moderate | moderate | good |
Protan and deutan are hard to separate, and we say so. Their confusion axes sit only about 15° apart in color space, so a run that clearly finds a red-green pattern often still cannot tell you which of the two it is. When the evidence does not clearly favour one, the result says “protan or deutan” rather than guessing. Telling them apart properly needs an anomaloscope.
Your retina carries three kinds of color-sensitive cone cell, each most responsive to a different band of wavelengths: L (long, reddish), M (medium, greenish) and S (short, bluish). No cone sees “red” or “green” on its own — each simply reports how strongly it was stimulated.
Color is what your brain computes from the ratios between those three numbers. This is why color is a perception rather than a property of light, and why a screen can fake almost any color with just three primaries: it only has to produce the same three cone responses that the real thing would.
Before those signals leave the eye they are recombined into opponent channels — roughly red-versus-green, blue-versus-yellow, and light-versus-dark. That reorganization is why color deficiencies come in the pairings they do, and why this test measures error along those same axes rather than along raw red, green and blue.
Prevalence figures are for populations of northern European descent, where red-green deficiency is most common; rates differ meaningfully between ancestries.
“Colorblind” is nearly always a misnomer. Total absence of color is vanishingly rare — the usual case is one cone type shifted or missing, which compresses part of the spectrum rather than erasing it.
Why it mostly affects men. The genes for the L and M cones sit on the X chromosome. Someone with one X has no second copy to compensate if it carries the variant, so a single affected X produces the deficiency; someone with two X chromosomes usually has a working copy to fall back on. The S cone gene is not on the X chromosome, which is why tritan deficiency does not follow the same pattern.
Switch between deficiency types and drag the severity. Watch which swatches slide into each other — those collapsing pairs are exactly what both games use as their hardest distractors. Then tick the correction box to see what the in-game assist does about it.
The simulation uses the Machado, Oliveira & Fernandes (2009) model, which represents partial deficiency at any severity rather than only the complete form. It shows a reasonable approximation on an uncalibrated screen — not what any particular person sees.
Color-correcting glasses are notch filters. They remove a narrow band of wavelengths where the L and M cone responses overlap most, which widens the gap between the two signals your brain is comparing. Colors that used to sit almost on top of each other separate a little.
They add no information. If a cone type is entirely missing, there is no overlap to sharpen and the glasses do essentially nothing — they help some anomalous trichromats, not dichromats, and they need bright light to work at all.
A screen has an advantage a lens does not: it can change the color before emitting it. The assist re-maps each color so that the differences you cannot resolve are re-expressed as differences you can — a technique called daltonization.
Grays are left exactly untouched, so nothing takes on a color cast. Measured on a confusable pair, an affected viewer sees the gap widen by around 3× for protan, 13× for deutan and 5× for tritan at the default strength.
Neither is a cure, and assist mode disables the test. These techniques rearrange colors so more of them are distinguishable; they do not restore a missing cone or make you see a color you have never seen. And because the assist changes the colors before they reach your eyes, a run played with it on cannot measure your color vision — so those runs report the score and skip the estimate rather than reporting a number we know is meaningless.
Panel type, brightness, color profile, night-shift and blue-light filters all shift what is actually emitted. A real clinical test controls the light source; this cannot.
Ambient light changes how a screen looks. Clinical testing uses a standardized illuminant for exactly this reason.
Even 50 rounds is a small sample. The confidence tier reflects this honestly, and its ceiling is “good” — never “diagnostic”.
Tiredness, distraction and rushing all widen your errors. The test detects the extreme case and calls it unreliable, but it cannot read your state of mind.
Their confusion axes are close enough that a clear red-green finding often still cannot say which one. Separating them properly needs an anomaloscope.
“No strong signal” means this test found nothing, not that nothing is there. Mild deficiencies can pass unnoticed here.
This is a game, not a medical device. Nothing here is a diagnosis and nothing here should be used to make a decision about your health, your license or your job. If anything you see here matters to you, or if your color vision has changed recently, see an optometrist or ophthalmologist — a sudden change in color vision can signal a condition worth catching early. Proper testing uses Ishihara plates, the Farnsworth-Munsell 100 hue test, or an anomaloscope.
Yes, completely. No account, no download, and no tracking other than the web host's analytics. Everything runs in your browser, and your results never leave your device. You can help me out, though, by donating or using my Amazon links to make purchases. Those links give me a small commission at no cost to you (other than your purchase), and it helps keep the site running. Even if you do not purchase what is advertised, you can support the site by using my links to make your own purchases on Amazon. Not all purchases will give me a commission, but many will. Thank you for considering it, and for playing!
Deuteranomaly, by a wide margin. It is a shifted rather than missing M (green) cone, and it accounts for roughly 6% of men on its own — about three quarters of all color-vision deficiency. Full deuteranopia, protanomaly and protanopia together make up most of the rest. Tritan deficiency and true monochromacy are rare enough that a tritan result here is worth confirming rather than believing outright.
The everyday signs are ordinary rather than dramatic: arguing about whether something is green or brown, unripe and ripe fruit looking the same, struggling with red text on a dark background, or a status light you cannot read at a glance. Most people with a mild deficiency never notice, because they learned the names for the colors they see and nobody ever contradicted them. Both games here are a low-effort way to find out whether your errors have a direction — and an optometrist is the way to find out for certain.
No. Ishihara plates are fixed printed images and a plate test asks whether you can read the number hidden in the dots. This measures the direction of your errors across many freshly generated rounds instead, which sidesteps two problems: the plates are widely available online and easily memorised, and a screen cannot reproduce their printed inks faithfully anyway. Ishihara remains the standard screening tool in a clinic. This is not a replacement for it.
Inherited color blindness cannot be cured. It comes from the cone photopigments your genes coded for, and no lens, filter, supplement or app changes that. Filters and daltonization can make more colors distinguishable from each other, which is genuinely useful, but neither adds a color you have never seen. Acquired color-vision loss is a different matter — that stems from an underlying condition, and treating the condition is what matters there.
For some people, some of the time. They are notch filters: they cut a band of wavelengths where the L and M cone responses overlap, which widens the gap between the two signals your brain compares. That can help an anomalous trichromat — someone whose cones are shifted but present. If a cone type is genuinely missing there is no overlap to sharpen and the glasses do essentially nothing, and most designs need bright daylight to do anything at all. They are worth trying with realistic expectations, ideally after an eye exam tells you which kind you have.
Yes, and that case matters more than the inherited one. Acquired color-vision changes can follow cataracts, glaucoma, diabetes, optic-nerve problems, certain medications, or an injury. The distinguishing feature is that something changed — inherited deficiency has been there your whole life. If colors look different to you than they used to, that is a reason to see an eye doctor promptly, and not something to work out from a browser game.
Yes — both games are built for touch and the sliders are designed to be draggable with a thumb. The usual caveat applies harder on mobile, though: phone screens auto-adjust brightness and color temperature constantly, and Night Shift or a similar warm-tint mode will skew the estimate. Turn those off for a run you intend to take seriously.
It can give you a reasoned, experimental estimate, and with 40 or more rounds that estimate is right most of the time in simulation. It is still not a diagnosis. Treat a positive result as a good reason to book a proper test, not as the answer.
Ten is a quick game but can only ever support low confidence. Twenty to thirty gives a moderate reading. Fifty is where the estimate is strongest. If you actually care about the result, play fifty with assist off.
Your answers were scattered so widely and so evenly that they carry no usable signal — which is what random clicking looks like. This is deliberate: a test that produced a confident-sounding verdict from noise would be worse than useless.
Because the assist changes the colors on their way to your eyes, and those colors are exactly what the estimate measures. A corrected run tells you how well the filter works, not how your color vision works. Play one run with assist off for a reading, then turn it on to see the difference.
It has the same goal but a different mechanism. Glasses filter light on the way in; this re-maps color before the screen emits it, which is a technique called daltonization. Both widen the gap between confusable colors. Neither restores a missing cone.
The quiz, slightly. Its wrong answers are chosen to be colors that collapse onto the target under one specific deficiency, so a single wrong pick carries more information than a near miss on a slider. Chroma Match makes up for it with a richer signal per round. Playing both is the best answer.
Only the analytics that the web host collects! Your difficulty, round count, assist mode and best scores are kept in your own browser's local storage, and clearing your site data removes them.