How a guess is scored
Every guess in Color Hunch is scored on how far your colour sits from the real one — not in the sliders you moved, but in how different the two look to a human eye. The measure is CIEDE2000, and this page explains what that means and why the distinction matters.
It matters because the obvious alternative is wrong in a way players can feel.
Why slider distance is the wrong measure
The simple way to score a colour guess is to compare the dial positions: how far is your hue from the target hue, your saturation from its saturation. It takes one line of code and it produces a game people quietly stop trusting.
The reason is that equal movements of a dial are not equal changes to the eye. A ten-degree shift in hue is glaring in the greens and nearly invisible in the deep blues. Move the lightness slider a little in a pale colour and everyone sees it; move it the same amount in a dark one and almost nobody does. Score on slider distance and the game punishes you for approaching from the wrong side of a hue while rewarding a landing that plainly looks wrong.
What a colour difference actually is
Colour difference is measured by placing both colours in a space built so that distance in the space matches difference to the eye, and then measuring the distance between them. That number is called ΔE — delta E, where E is for Empfindung, the German for sensation.
The first such formula, CIE76, was a straight line between two points in Lab space. It was a real advance and it had known weak spots: it overstated differences in saturated colours and understated them in blues. CIEDE2000 is the correction, adding weighting terms for lightness, chroma and hue, plus an interaction term for the blue region where the earlier formula did worst.
What Color Hunch does with it
When you submit a guess, the colour you dialled and the colour that belongs there are both converted from sRGB to Lab, and CIEDE2000 gives the distance between them. That distance becomes the round's score, and the stars for the stage come from the scores of its rounds.
The number you see is a closeness percentage, not the ΔE itself. That is a deliberate choice: a ΔE of 4.2 tells you nothing you can act on, while 'Closeness 87' tells you the same thing in a form you can compare with your last attempt. The ΔE stays where it belongs, in the scoring.
Why the same formula runs everywhere
The app, the game server and this website all score with the same implementation, checked against the same reference values. A guess that scores 87 in the browser scores 87 in the app.
That sounds obvious and is easy to get wrong: three codebases in three languages will drift apart unless something holds them together. Here it is a test with fixed input pairs and expected outputs — if a port disagrees with the reference, the build fails rather than the leaderboard quietly becoming meaningless.
What ΔE values mean
As a rough guide, this is what a given colour difference looks like to someone with normal colour vision, under decent light:
| ΔE | What you would notice |
|---|---|
| Under 1 | Not distinguishable by eye. The two colours read as the same colour. |
| 1 to 2 | Visible to a trained eye on close inspection, usually side by side. |
| 2 to 10 | Noticeable at a glance. Most people would call these two different shades of the same colour. |
| 11 to 49 | Clearly different colours that are still related — a red and a slightly orange red. |
| 50 and above | Different colours entirely. Nobody would confuse them. |
These bands are a convention rather than a specification: they are how the field talks about the numbers, and different sources draw the lines in slightly different places. Treat them as a way to read a score, not as a standard.
CIEDE2000 is defined by the CIE (Commission Internationale de l'Éclairage) in CIE 142-2001. The implementation here follows Sharma, Wu and Dalal (2005), 'The CIEDE2000 Color-Difference Formula', which is the paper most implementations are checked against and which documents the test data used for that check.