A grid of identically colored tiles, one of them subtly different. Find it and you advance: the grid gets denser and the color gap narrows.
It looks like a simple game, but it's hard to tell what counts as good. This post works out how different the colors actually are at each stage, and where those numbers meet the limits of human vision.
1. What the test actually measures
One thing first: this is not a hue test.
The odd tile takes the others' color and raises red, green, and blue by the same amount. Move all three channels together and the hue stays put — only the brightness shifts.
Other tiles: rgb(120, 90, 160)
Odd tile: rgb(137, 107, 177) ← all three +17
So what's being measured is your ability to separate similar brightness levels, not to tell red from orange. For the same reason it cannot be used as a color blindness test. Color vision deficiency is about distinguishing specific hues, and there is no hue difference here at all. That needs a purpose-built test such as the Ishihara plates.
2. How the difficulty climbs
Two things tighten at once: the grid gets denser and the color gap shrinks.
| Stage | Grid | Tiles | Color gap (RGB) |
|---|---|---|---|
| 1 | 2×2 | 4 | 45 |
| 3 | 3×3 | 9 | 36 |
| 5 | 4×4 | 16 | 29 |
| 8 | 6×6 | 36 | 21 |
| 10 | 7×7 | 49 | 17 |
| 12 | 8×8 | 64 | 14 |
| 16 | 8×8 | 64 | 9 |
| 20 | 8×8 | 64 | 6 |
| 24 | 8×8 | 64 | 3 |
| 27+ | 8×8 | 64 | 2 |
The gap drops by 10% per stage. Starting at 45, it's down to 6 by stage 20 — 2.4% of the 0–255 range.
The grid stops growing at stage 12 (8×8, 64 tiles). After that the board never gets busier; only the color gap keeps closing.
3. 🔴 Past stage 27 it stops getting harder
This is the real structure of the test. The color gap has a floor of 2.
Reach stage 27 and the difficulty is fixed from then on. Stage 28 and stage 40 are the same problem: find a tile that differs by RGB 2 among 64.
So records past stage 27 come from searching faster, not seeing better. The timer starts at 15 seconds and a correct answer returns 1 second — but a wrong tap costs 2. One mistake takes two correct answers to undo, which makes the late game a contest of search speed and error control rather than eyesight.
For scale, an RGB gap of 2 is two steps out of the 256 an 8-bit display can show — about 0.8%. That sits close to the limit of human brightness discrimination, and past it your screen usually gives out before your eyes do.
4. Which is why scores don't compare
The same person with the same eyes will score differently depending on the setup. The late-stage differences are so fine that everything hinges on whether the display can render them at all.
Screen brightness — dim screens crush fine brightness steps. Running on battery saver costs you stages outright.
Night mode / color temperature — warming the screen suppresses the blue channel and narrows the range. Leave it on and you're handicapped.
Panel quality — some budget panels simulate 8-bit with 6-bit plus dithering. On those, an RGB gap of 2 may render as literally the same color. That's not your eyes; the screen cannot produce the difference.
Ambient light — in a bright room, reflections bury the darker gradations.
So the absolute score means little, and only changes within the same setup mean anything. Scoring below a friend more likely means a different screen than worse eyes. It's exactly the problem the Reaction Time Test has with refresh rate and input lag folded into the number.
5. One more thing — bright draws are easier
There's a quirk worth knowing. The base color is drawn randomly per channel in the 20–219 range, and the odd tile adds the gap on top — but it cannot exceed 255.
So when the base comes out bright (say 219), adding 45 doesn't reach 264; it clips at 255 and the real difference is only 36. A round with a bright base is slightly easier than one with a dark base.
This only happens in the early stages, where the gap is large; later on it barely matters. But it does explain why some rounds feel unusually easy at the same stage.
6. To give yourself a fair shot
- Brightness to maximum, night mode off. This is the biggest variable
- Play in a dark room. Reflections erase fine gradations
- Look straight on. Laptop panels in particular lose gradation off-axis
- Don't scan tile by tile — soften your gaze over the whole grid and let the odd one pop. Peripheral vision is actually good at brightness differences
- Remember that a miss costs 2 seconds. Taking one more beat beats guessing
To try it, here's the Color Discrimination Test. It starts at 15 seconds, returns 1 second per correct answer and takes 2 for a miss, so the late game becomes a race against the clock.
For reflexes, the Reaction Time Test is the neighboring tool; for visual search speed, try 1 to 50.
Summary
- The test measures brightness, not hue — all three channels move together. It is not a color blindness test
- The gap shrinks 10% per stage, reaching RGB 6 (2.4%) by stage 20; the grid locks at 8×8 from stage 12
- 🔴 From stage 27 the gap is pinned at 2 and stops getting harder — past that it's search speed and error control, not eyesight
- Brightness, night mode, and panel quality drive the score. Don't compare with others; compare your own runs in the same setup
- A bright base clips at 255, which makes some early rounds easier
