← Back to game

About Colourdle

See a colour. Guess its name. Score by how close your guess is to the actual name, and how close the colour you named is to the target.

Game Options

Two independent choices shape your game:

Timing:

Game type:

These combine freely: Daily Classic, Daily Explore, Freeplay Classic, Freeplay Explore. Each Daily combination has its own puzzle and saved results.

Palette Filter

Choose a colour source to play with:

When a palette is selected, both the target colours and your guess matching are restricted to that source. Each palette has its own daily puzzle.

Scoring

Each round produces two scores:

Name score = ((1 − rank / pool_size) ^ 2) × 100
Colour score = 100 / (1 + (Delta E / 30)^1.8)

The two scores are combined using a power mean (p=3), which favours the higher of the two values. This means doing very well on one axis partially compensates for the other.

Final = ((name^3 + colour^3) / 2) ^ (1/3)

The maximum base score per round is 100. If your guess matches the exact colour entry (same hex value), your round total is doubled as a bonus — so scores above 100 per round are possible but not expected.

How Matching Works

When you type a guess, the game finds the closest colour name in the active pool using POEM (Pareto-Optimal Embedded Matching), adapted from Brereton et al. 2020. Instead of relying on a single matching algorithm, POEM combines six independent similarity measures and uses Pareto dominance to find the best consensus match.

The six measures are:

A colour name is only chosen if no other candidate beats it on every measure simultaneously. This means "gray sky" correctly matches "Sky Gray" even though no single measure would reliably pick it. The result is robust fuzzy matching with zero dependencies — no ML models, no API calls, just cheap string metrics that are smart together.

See it in action — compare Colourdle's adaptation against the original paper's fitness-based selection, side by side.

This means there are two strategies: guess the exact name (high name score), or guess a colour that's perceptually close (high colour score). The best guesses do both.

Technical Notes

Credits

Created by Drew Brereton. Source code on GitHub.

Colour data sourced from Crayola, Pantone (via community datasets), XKCD colour survey, RAL colour standards, and meodai/color-names — a community-curated collection of over 30,000 handpicked colour names.