Colourdle uses POEM (Pareto-Optimal Embedded Matching) to find the closest colour name to your guess. The original algorithm was published in Brereton et al. 2020 for drug property prediction. Colourdle adapts the core idea — multiple distance measures combined via Pareto dominance — but uses a different selection step tuned for nearest-match lookup. Type any colour name below to compare both approaches.
Both algorithms compute six string distance measures for every colour name in the pool (~3,300 colours), then use Pareto dominance to find a consensus best match without weighting or tuning any individual measure.
Colourdle prunes to the top 50% per measure, finds the strict Pareto front (colours not beaten on every measure by any other), then picks the candidate with the lowest mean distance. Fast and focused on proximity to the query.
Paper computes pairwise dominance between all candidates with relaxed tolerance (a candidate can claim dominance even if ≤10% of measures disagree). Each candidate gets a fitness score:
Fitness = MeanDominance × (NumDominating + 0.05) / (NumSubmitting + 0.05)This favours candidates that dominate many others overall — ideal for classification tasks where you want the most representative neighbour, but sometimes picks a globally dominant candidate over the one actually closest to your query. Try "gingerbread" for an example where the approaches disagree.