How to Play

Study the colour breakdown

The pie chart shows the exact colour split of today's flag, measured pixel by pixel from the real design. A flag that's mostly red and white points somewhere very different to one that's green, gold and black.

Type your guess

Enter any country name, or pick one from the dropdown. It has to be a real country to count as one of your five guesses.

Unlock a hint with every miss

Each wrong guess reveals the next clue, in order: neighbouring countries, population, size, then continent. Your fifth guess is the last chance, no new hint after that.

See the answer

Get it right and you win, however many guesses it took. Run out of guesses and the real flag and country are revealed.

Colour Accuracy

Both your drawing and the real flag get converted into the CIE Lab colour space, a model built to represent colour the way the eye actually perceives it, rather than the raw RGB values a screen uses. Your drawing gets scanned into a colour histogram, grouped into clusters, and any cluster covering at least 1% of the canvas counts as a genuine, deliberate colour choice rather than edge noise. Each of the flag's real colours then gets greedily matched to its single closest cluster in your drawing, one real colour, one matched cluster, best pairing claimed first.

For each matched pair, the distance between them isn't a flat Lab difference. It's decomposed into three separate components, lightness (L), chroma (C), and hue (H), the same decomposition CIE94-style colour difference formulas use, and hue is weighted more heavily than lightness. That's deliberate: a lighter or darker version of the correct colour is far more forgivable than a genuinely different hue, so navy standing in for sky blue costs you less than orange standing in for red, even if the raw numerical distance were identical.

That weighted distance gets converted into a 0-100 score per colour, and each colour's contribution to your final Colour Accuracy score is weighted by how much of the real flag it actually covers, getting a colour that's only 3% of the flag slightly wrong barely matters, getting the dominant colour wrong matters a lot. On top of that, any cluster in your drawing that was significant enough to count but didn't get matched to any real colour, paint that doesn't correspond to anything on the actual flag, is treated as wasted canvas space and pulls the whole score down as a multiplier, not just a rounding error.

Colour Proportion

This reuses the exact same cluster-to-colour matching from Colour Accuracy, not a fresh comparison, so the two scores are always talking about the same matched pairs. For each match, it compares what percentage of your drawing that cluster actually covers against what percentage of the real flag that colour actually occupies. This is purely about the split, not the shade, you could pick the perfect red and still lose points here if you drew far too much of it or far too little relative to the rest of the flag.

Placement

Your drawing and a rendered version of the real flag get compared pixel by pixel, at identical dimensions, using Intersection over Union (IoU), a standard way of measuring how well two regions actually overlap, borrowed from image-segmentation work. For each official colour, it's the number of pixels where your drawing and the real flag agree on that colour, divided by the total number of pixels where either one has it. A perfect overlap scores 1, no overlap at all scores 0, and the per-colour scores are averaged into your final Placement percentage. This is what actually catches a rough blob of the right colour sitting over a detailed emblem, the colour might be there, but the pixels won't line up.