What this test measures
One tile in the grid is a slightly different shade from the rest. Your whole job is to find it and click it. The grid starts at 2×2 and grows every three levels until it reaches 6×6. Meanwhile the difference between the odd tile and the rest keeps shrinking. Early rounds make the odd tile about 13 points lighter or darker. By the late levels the gap can be as small as 2 points, and it never goes lower. You get three lives, or just one in hard mode, which keeps its own record. A wrong click costs one, and your score is the last level you cleared.
Two different abilities decide how far you get. The first is discrimination, the smallest lightness difference your eye can resolve, which vision researchers call a just-noticeable difference. The second is visual search. On a 6×6 board you are checking 36 tiles. A difference you would spot instantly between two side-by-side squares can hide in a crowd. The hue is randomised every round, so you cannot memorise one color and coast. Each level is a fresh comparison.
Why there is no percentile here
Only one test on this site shows a percentile — sequence memory, which has a published norm behind it — and this one does not. Nobody has published a normative distribution for a browser tile-discrimination game of this design, so any percentile we showed you would be invented. There is a deeper problem too: the thing being measured is not cleanly you. Whether you can tell two shades apart at a 2-point difference depends on your screen, its brightness, its color settings and the light in your room. Those things swing the result more than the player does. A percentile here would mostly be ranking monitors.
Real eye-clinic tests, like the Ishihara plates, earn their numbers by controlling what a web page cannot: printed colors viewed under standard lighting. Without that control, comparing one person's score with another's is noise dressed up as data. The comparison that does mean something is you against yourself, on the same device, under the same conditions.
Your screen is part of the test
The same person can score several levels apart on two devices in the same afternoon. Here is what changes underneath you:
The panel itself decides whether a difference exists at all. A decent IPS or OLED panel driving 8 bits per channel renders a 2-point lightness step as two genuinely different colors, while a cheap 6-bit panel may round both tiles to the same output. When that happens no amount of staring helps, because the difference never reaches the screen.
Brightness matters next. A dim screen compresses shades together, especially darker ones, and raising brightness stretches them apart again. Night shift and blue-light filters pull the whole display toward orange and can flatten the very differences you are hunting for, most noticeably on cool hues. A normal dark theme leaves the game's colors alone, but "smart invert" and forced dark mode on some phones repaint web pages, and that can distort the tiles.
Then there is the room and the angle. A window behind you lays a sheet of glare over every tile, and reflections raise the black level of the screen and eat subtle contrast. On cheap panels colors also shift as your eye moves off-axis, so a corner tile reaches your eye at a slightly different angle than a centre tile. That alone can fake a difference or hide a real one.
This is also why your phone score and your laptop score rarely agree. The two devices have different panels and different automatic behaviors, and a phone's auto-brightness can quietly change mid-game when a cloud passes the window.
Color vision, and who this test excludes
Around 1 in 12 men and 1 in 200 women have some form of color vision deficiency. The other tests on this site never use color as the only signal. Shape, position, motion or text always carries the message too. This test is the one exception, and the reason is simple. Color discrimination is the thing being measured, so there is nothing else to fall back on. If you are color-blind, some rounds of this game will be unfair to you in a way the other tests are not. We would rather say that plainly than pretend otherwise.
What this test cannot do is diagnose anything. A low score here tells you nothing reliable about your eyes, because a bad screen produces exactly the same result. If you think you might be color-blind, the real tests are the Ishihara plates and the Farnsworth-Munsell test, given by an eye doctor under proper lighting. A browser game is not a substitute for an eye examination, and this page should never be treated as one.
How to see the difference for longer
Sweep the grid in a fixed order, row by row, rather than letting your gaze bounce around. Staring hard at one spot backfires. Your eye gets used to the shade you are staring at, and after a few seconds the whole board blurs into sameness. Keep your eyes moving and let the odd tile pop out in peripheral vision, which is surprisingly good at catching a mismatch.
Work the borders. Two shades that look the same from a distance are easier to tell apart along the line where two tiles nearly touch. Your eyes compare neighbors far better than memories.
Set the stage before you start. Turn off night mode, raise your screen brightness, face the screen straight on, and get glare off the panel. If the tiles have all started to look the same, look at something far away for ten seconds. Your eyes reset, and the differences come back.
And accept the ceiling honestly. Once the game reaches the 2-point floor, whether you can continue is decided largely by your display. Technique buys you levels in the middle of the run, not at the end of it.