The color game guide to remembering a color
What is the color game?
The color game is a test of one narrow ability: holding an exact color in your head for a few seconds. A color fills the card, a countdown runs, and the color vanishes. Three sliders appear, for hue, saturation and brightness, and the panel behind them shows whatever you have mixed so far. When it matches what you remember, you lock it in. The original comes back beside your version, each labeled with its three values, and a score from 0 to 10 tells you how close the two look.
That is the whole loop. There is nothing to name, no palette to learn, no hex codes. Five colors make a color game run, scored out of 50, and the game keeps your best and your average for each difficulty on the device you play on. The daily color game deals the same five colors to everyone on the date; unlimited runs deal a fresh five as often as you like.
What makes the color game interesting is not the rules but the gap it exposes. While the color is on screen it feels completely known. Two seconds after it is gone, you discover that what you kept was a word and a mood, and that the sliders ask for far more than that. The rest of this page is about why, and about how to close the gap.
Why colors are so hard to remember
Psychologists have run a version of the color game in the lab for over twenty years, and their results explain most of what happens on the sliders.
Memory keeps the name, not the shade
In 2015 Gi-Yeul Bae and colleagues asked people to reproduce colors on a continuous wheel, some while the color was still visible and some from memory. Answers from memory were pulled away from the borders between color names and toward the center of each category, by up to ten degrees of hue, and even a sub-second delay roughly doubled the pull compared with matching a color in view. The brain seems to store a color as a label, "blue", plus an estimate of the exact shade, and when the estimate fades the label takes over. A teal drifts toward plain green or plain blue; a mauve becomes pink or purple. Every color game round replays this experiment.
Berlin and Kay counted the basic color terms of English in 1969 and found eleven: black, white, red, green, yellow, blue, brown, orange, pink, purple and gray. Pointer and Attridge estimated in 1998 that the number of surface colors a person can tell apart is about 2.3 million. Eleven labels for millions of colors is the whole problem of the color game in one ratio.
Remembered colors come back more vivid
In 1960 C. J. Bartleson asked fifty people to pick the colors of ten familiar things, grass, sky, brick, skin, from a large set of Munsell chips, from memory. Every one of the ten came back different from measurements of the real thing, and in most cases more saturated and lighter. Memory does not just blur a color; it exaggerates it. In 2006 Hansen and colleagues showed that this leaks into perception itself: asked to adjust a photograph of a banana until it looked gray, people stopped on a slightly bluish banana, because a truly gray one still looked faintly yellow to them.
For the color game sliders, this has a direct consequence. If your rebuild keeps coming out as the more vivid of the two on the reveal, you are not misremembering by accident. You are doing what everyone does, and the fix is to aim below what feels right.
Two or three colors, then nothing
Zhang and Luck, in Nature in 2008, used a wheel of 180 equally bright colors in CIELAB, the same space the color game scores in, and found that people hold a precise copy of about three colors and nothing at all of the rest. A year later they showed that a held color does not fade gradually; it stays precise for a while and then drops out whole, which they called sudden death. The color game asks for one color at a time, well inside that limit, so a bad round here is almost never a color forgotten outright. It is a color that was held and then drifted, and the drift has a direction you can learn.
How the color game scores a rebuild
Slider numbers would make an unfair score. Hue 359 and hue 1 are the same red. At zero saturation every hue is the same gray, and at zero brightness every color is black. So the color game scores what you would see, not what you typed.
Both colors are converted from hue, saturation and brightness to sRGB, then to CIELAB, the CIE's 1976 color space built so that equal distances stand for roughly equal visible differences. The gap between the two points is measured with CIEDE2000, the CIE's color-difference formula of 2001, written ΔE00, which corrects the raw distance for the way the eye is more forgiving of lightness errors on very dark and very light colors, less forgiving of hue errors on strong colors, and oddly tolerant in the blues. A ΔE00 of 0 is an exact match. About 1 is the smallest difference most people can pick out. Dentists, who match tooth shades for a living, have measured where visible begins: in a 2015 study across seven sites, half of 175 observers could see a difference of 0.8, and half would accept 1.8 as a match. Ten is plain side by side. Thirty is another color.
The color game score follows a smooth curve from that distance. A ΔE00 of about 4 scores roughly 9.5. About 10 scores roughly 8. A ΔE00 of 25.25 scores exactly 5, and the curve is built so that it is gentle on near misses and steep further out, which means the right color a shade off still scores well while a different color scores little. Two small corrections sit on top. When the hue is within about 25 degrees of the answer on a reasonably saturated color, up to a quarter of the missing points are given back, because getting the hue right on a strong color is the hard part. When the hue is more than 30 degrees off on a saturated color, up to 15 percent of the score is taken away. Finally the second decimal is nudged by a few hundredths in a fixed way, so that two different near-misses do not show identical numbers.
The color game's implementation is checked against test pairs published by Sharma, Wu and Dalal in 2005, which is the standard way to confirm that a CIEDE2000 routine is correct. The five scores in a run are added two decimals at a time, and the run total is out of 50.
Hue, saturation and brightness, explained
Three numbers describe any color on a screen, in the order a painter thinks about mixing one, and they are the three sliders of the color game.
Hue is which color it is, as a position on the wheel: red at 0 degrees, yellow at 60, green at 120, cyan at 180, blue at 240, magenta at 300, and round to red again at 360. The hue slider in the color game runs through the whole wheel.
Saturation is how far the color is from gray, from 0 percent, which is gray at that brightness, to 100 percent, the purest version of that hue the screen can produce. A dusty rose and a hot pink can share a hue and differ only here.
Brightness is how much light the color has, from 0 percent, black, to 100 percent, as much as the screen gives that hue. It is not lightness. On the CIELAB lightness scale, where 0 is black and 100 is white, full-brightness yellow sits near 97 and full-brightness blue near 32, because a screen emits far more light for yellow than for blue. So a deep, dark-looking blue or violet can be at full brightness, while a yellow turns olive well before its brightness slider is halfway down. This is why full blue looks so much darker than full yellow at the same setting, and why the color game's brightness slider rewards a different instinct for every hue.
Splitting a color into these three goes back to the painter Albert Munsell, whose A Color Notation of 1905 called them hue, chroma and value and arranged them in steps meant to look even. The sliders are the screen-age cousin of that system, HSB, also called HSV. Alvy Ray Smith wrote the conversion between it and the red-green-blue values a monitor actually uses for a paint program in the 1970s, and published it in 1978. The three interact: at zero saturation the hue slider does nothing, and at zero brightness nothing does anything, so a hue error costs little on a pale or dark color game round and a lot on a vivid one.
Easy, Hard and Brutal in the color game
The three difficulties change how long you can look and what gets in the way. They do not change the colors, which the color game draws from the whole wheel with saturation and brightness anywhere between 15 and 100 percent, and they do not change the scoring.
Easy shows each color for 5 seconds, with no clock on the sliders. This is where most people should start the color game, and where the habits below are easiest to practice, because five seconds is long enough to name the color and then name the lean.
Hard shows each color for 2 seconds. That is enough to read the hue and get a feel for brightness; it is rarely enough to judge saturation carefully, which is why Hard color game scores are usually a few points lower on exactly the colors that have no name.
Brutal shows each color for 2 seconds while wrong color words flash over it every half second, each in an ink chosen to contrast with the swatch, sometimes rotated, sometimes stacked three high, sometimes replaced by a plain block of another color. Then the sliders get 3.5 seconds, and when the clock runs out the color game locks in whatever you have. Brutal is less a memory test than a test of attention under interference, and a color game run above 35 on it is a genuine achievement.
The Stroop effect, and why Brutal is brutal
In 1935 John Ridley Stroop printed color words in mismatched inks, the word "red" in green, and timed people naming the ink. They were markedly slower than when naming the color of a plain block, and they made more errors. Reading is automatic for a literate adult, and the automatic reading of the word competes with the deliberate naming of the color. The effect is one of the most replicated in psychology.
The color game's Brutal mode runs that experiment on you while you are trying to encode the swatch. The words are chosen so that none of them is the plain name of the color you are looking at, and their inks are chosen to be readable against it, so the interference is maximal: every half second a loud, wrong label arrives through the channel your memory would most like to use. The defense is the same one that works in the lab. Do not read. Fix on the edge of the swatch where there are no words, decide the hue first, then brightness, and accept that saturation on a Brutal color game round is a guess you will refine over many runs.
How to get better at the color game
The experiments above each point to a habit. These five turn a color game 30 into a 40.
Name the color, then name the lean
The label is the part memory keeps best, and it drags the rest of the color toward a typical example of that label. So use the color game countdown to say, silently, how the color differs from its name: a blue that leans green and is a little gray; an orange that is nearly brown. The lean is exactly what the label would otherwise erase, and three or four words of it survive the gap far better than an unspoken impression.
Spend the seconds on saturation and brightness
Hue is the easy thing to hold. It has a name. Brightness and saturation are the two things a name says nothing about, and they are where most color game points are lost. Once you have the hue, give the rest of the countdown to the questions how dark is this, and how gray is this, and try to attach a rough number to each: about 70 percent bright, about half saturated.
Set saturation a notch low
Two biases push the same way. Remembered colors are more saturated than the originals, and memory pulls an in-between color toward a plain, typical one. If the color game reveal keeps showing your color as the more vivid of the pair, aim one notch below what feels right. Most players need this correction on almost every muted color and never on the vivid ones.
Work the sliders in one order
Hue first, then brightness, then saturation, then one last look at hue. A fixed order stops you chasing one slider while another is wrong, which is the most common way to turn a good memory into a bad color game score. On a pale or dark color, the hue matters least, so if time is short on Brutal, settle brightness before you touch anything else.
Read the reveal, not only the score
The color game reveal shows both colors with their values. Note which slider was off and which way: too light, too vivid, too warm. Most players miss in the same direction nearly every round, and a bias you know about is one you can correct on the next color. The results screen lists all five pairs for the same reason.
Daily colors or an unlimited run
The daily color game deals five colors from the date, so everyone who plays on the same calendar day sees the same five in the same order, including the same starting positions on the sliders. The set changes at midnight in your own time zone. You can play it on any difficulty, and the color game records one result per day on your device, with the difficulty noted in the share text so a friend on Brutal is not compared with you on Easy.
Unlimited color game runs deal a fresh five every time. They are the practice mode, and the stats screen keeps a separate best and average for unlimited runs on each difficulty. The numbers you want to watch are the averages. A best score is a lucky set of nameable colors; an average that climbs is a habit forming.
Reading your color game results
Each color scores from 0 to 10 with two decimals. A score of 8 or above means the two colors would get the same name and sit only slightly apart side by side, a ΔE00 of roughly 10 or less. From 9.5 the pair is a near copy, a ΔE00 of about 4 or less. From 6 you are in the right family on a visibly different shade. Under 4 the two look like different colors.
For a color game run, 30 is a fair first result on Easy and the number most people settle at before they start correcting their biases. A run of 40 means every color came back as the right color, a touch off. A run of 45 means you were within a couple of visible steps on each. The color game share text compresses this into five squares: green from 9.3, yellow from 7.5, orange from 5, red below. A row of five greens is the thing to chase.
Does the screen change the color game score?
Not the arithmetic. The color game compares the numbers it drew with the numbers you set, and those do not depend on the display. A dim phone, a night-light mode or a cheap monitor leaves the scoring exactly where it is.
What a screen changes is the difficulty. The original and your rebuild are shown on the same display, so a warm tint shifts both the same way and you can still match by eye. But a dim screen or a night mode squeezes dark colors closer together, which makes brightness harder to judge, and a very bright screen in a dark room exaggerates saturation. If you are comparing color game scores with a friend, play in similar conditions, or accept that the daily is a contest of eyes and screens together.
From Newton to CIEDE2000: how color became a number
The ruler the color game scores with took three centuries to build. Isaac Newton's prism experiments of 1672 showed that white light is a mixture of colors bent by different amounts, and his Opticks of 1704 bent the spectrum round into a circle, the first color wheel, with red and violet meeting where no spectral color exists.
Munsell's notation of 1905 gave every color three separate qualities in steps meant to look even, and his charts are still used to describe soil. In 1931 the International Commission on Illumination, the CIE, defined the color-matching responses of a standard observer, so that a color could be specified by numbers rather than by a painted sample. In 1943 Dean Farnsworth published the Farnsworth-Munsell 100 hue test, 85 colored caps to arrange by hue, still used to grade color vision.
In 1976 the CIE published CIELAB, a space built so that equal distances look roughly equally different. It was good but not perfect: distances in the blues and among very saturated colors did not match what people saw. The CIEDE2000 formula of 2001, developed by Luo, Cui and Rigg, added corrections for lightness, chroma, hue and a rotation term for the blues, and became the standard way to put one number on how different two colors look. Sharma, Wu and Dalal published a set of test pairs in 2005 so that anyone implementing it could check their work. That formula, checked against their pairs, is what turns a color game rebuild into a score.
Color game, colour game: one word, two spellings
Color reached English in the Middle Ages from Anglo-French, from the Latin color, which in Old Latin meant a covering. Colour was the usual English spelling for centuries, and scholars tidying words back toward their Latin roots made color an alternative. Noah Webster's dictionaries of 1806 and 1828 settled color for the United States, while British usage kept colour. Searches for the color game and the colour game land on the same page here, and the color game is the same under either spelling. The CIE itself, which defines the formula that scores it, writes colour.
The color game beyond the browser
The lab task the color game copies is called delayed estimation. Wilken and Ma introduced it in 2004: flash a few colored squares, blank the screen, and ask people to click the color of one of them on a wheel of 252 hues. Answering on a continuous scale instead of saying same or different turns a memory into a number, how many degrees round the wheel the answer landed from the truth, and that number is what let Zhang and Luck measure the two-or-three-item limit. The color game's sliders are the same idea with three dimensions instead of one.
Outside the lab, people are paid to do what the color game asks. Paint matchers at hardware stores, dental technicians choosing a tooth shade, printers checking a proof against a swatch, and colorists grading film all hold a color across a gap of seconds or minutes, and all of them learned the habits above. If you would rather test colors you have carried for years than for seconds, try remembering the exact color of a stop sign or a banana before you look at one. Bartleson's subjects could not do it in 1960, and nobody has done much better since.
