Name that hex code

Given a color as a hex code (#fffe40) and four names from the xkcd color survey, how often does Jev pick the survey's name, and does it fall for the nearest similar color?

result146 questions

Given a color only as a hex code like #fdff63, Jev picks the name people gave it in the xkcd color survey 71% of the time, where guessing would get 25%. Every miss lands on the nearest similar shade, never on a far color. Raw hex codes are a weak spot TypeSafe already documents; this puts a number on it.

the survey's name71%

nearest similar color29%

a far color0%

How to read this: Three bars add up to Jev's picks: the survey's own name for the color, the nearest look-alike shade, and the two colors that are nothing like it.

146 colors; 90% interval [0.637, 0.774].

In short

Shown only a hex code and four names, Jev picked the name the xkcd survey crowd used for 71% of 146 colors, against 25% by chance.
Every miss went to the closest look-alike shade and none to a far color, so Jev gets the color family right and struggles with the exact shade.
The xkcd list ships with common software, so part of this may be memory of name-code pairs rather than working out the color.

What the data shows

Did you mean? meme: #fdff63 is light mustard; canary

Jev picks the survey's name 71% of the time (90% interval 64% to 77%), far above the 25% of guessing.
Every miss is a near miss. The other 29% all go to the look-alike shade; not one pick goes to a far color.
So Jev never confuses a yellow with a brown, but it often can't tell canary from light mustard.

What it means, and what it doesn't

Jev knows roughly what color a code is. It struggles with the fine distinctions, which is where names like "canary" and "light mustard" actually differ. For design work that means: trust it for the family, check the shade.

It doesn't mean Jev "sees" colors from codes; some of this may be memory of the widely copied xkcd list. And it confirms a known limit rather than revealing a new one.

Caveats

A known weak spot. TypeSafe lists raw numbers, and hex and RGB values especially, among Jev's documented weak spots. This experiment measures that limit rather than discovering it.
It may be recall, not seeing. The xkcd color list is copied everywhere: the popular plotting library matplotlib ships all 949 names with their exact hex codes. Jev may partly be remembering name-code pairs from code it was trained on rather than working out the color.
The look-alike can be a fair answer. The near option is the closest other survey color, at least 40 steps away on the red-green-blue scale. Some of those pairs are genuinely hard for people too, so a "miss" to the look-alike is often a defensible pick.

Jev on this experiment

Would a person find it interesting to read?: Yes72%
Does it describe you?: Yes53%
Would you have predicted it?: No59%
How fair is the comparison?: The comparison is shaky
How much should a reader rely on it?: A little
Which caveat matters most?: It may be recall, not seeing88%

Why ask this

Designers, developers and anyone working with a web page constantly deal with colors written as codes: #fdff63 is a pale yellow. A model that helps with that work needs to know roughly what color a code is, or at least which name fits it.

TypeSafe's own documentation says Jev is weak at raw numbers like these. This experiment measures how weak, with an answer key made by a huge crowd, and separates "no idea at all" from "right family, wrong shade".

How this was done

The people and the data

In 2010 the webcomic xkcd ran an online color survey: people were shown random colors and typed whatever name came to mind. About 222,500 people took part, and the result is a list of 949 colors with the name most people used for each (released into the public domain). It has since become a standard reference, built into common software.

The project drew 150 of those 949 colors. For each, Jev saw four names: the survey's own name, the nearest other survey color that is still clearly different (at least 40 steps away on the red-green-blue scale), and two colors far away. 146 of the 150 have results and are counted here.

What Jev was asked

Which name fits the color with the hex code #fdff63 best?
canary · tomato · deep brown · light mustard

(The survey's name is "canary"; "light mustard" is the look-alike.) Each question was asked with the names in three different orders, and the answers averaged.

How it was measured

How often Jev's top pick is the survey's name, against 25% for a random guess, with a 90% interval. For the misses, whether they went to the look-alike (right family, wrong shade) or to a far color (no idea).

Where these questions live

146 questions across 1 topic of the map. Each opens on the map with every question in it.

Color

Color Names146

Every question

All 146 questions behind this result, the telling ones first: the examples the analysis points to, then the ones where Jev misses, biggest gap first.

Jev’s own answer

Showing 0 of 146