Colordle Solver

The solver runs the same color math the game does. When Colordle scores your guess, it converts your color and the target into Lab color space, computes a Delta E difference (the CIEDE2000 formula, the same one image-editing tools use), and shows you 100 minus that difference as your percentage. Two colors that look nearly identical score in the nineties; opposite ends of the spectrum score near zero.

The solver inverts that process. Every candidate color in the named-color list gets the same Delta E computation against your guess, and any candidate whose computed score does not match the percentage you observed is eliminated. Not ranked lower. Eliminated. The tolerance is two hundredths of a point, because the game displays two decimals and the math has to land inside that window.

That exactness is the whole trick, and it is why I trust the output more than my own eye on a bad monitor. If the solver says a color cannot produce 34.71 against your guess, it cannot. Your display might lie to you; the arithmetic does not.

A single guess-plus-percentage pair does something geometric: it defines a shell around your guess. Every color sitting at that exact perceptual distance survives; everything closer or farther is out. On its own, one shell still leaves a lot of candidates, which surprises people who expect one guess to nearly solve the puzzle.

The second guess is where it gets unfair. A second shell, centered somewhere else, intersects the first one, and the surviving set is the intersection. In practice that means two decent guesses routinely cut thousands of named colors down to a list that fits on screen. Three guesses usually end the puzzle.

I think of it as triangulation, because that is essentially what it is. Each percentage is a distance measurement to an unknown point, and two or three distance measurements from different positions pin the point down. Sailors navigated by this for centuries; Colordle players can too.

Guess one, I play in the game itself: a named color chosen to split the space, something central rather than an exotic edge shade. The game hands me a percentage, say 19.69. I type the color name into the solver, select it from the dropdown, enter 19.69, and filter.

The candidate list that comes back is my palette for guess two. I pick the candidate furthest from my first guess — distance between guesses is what makes the second shell intersect the first usefully — play it in the game, and bring its percentage back. Two rows in, the list is usually short enough that I can read every remaining name.

Guess three is confirmation. If two candidates survived both rows, one of them is the answer, and the third guess settles it. If the list is somehow still long, the cause is almost always the same: my two guesses were too similar. The fix is not more guessing. The fix is a third guess chosen to be genuinely far from both.

Not all guesses filter equally. A guess at the extreme edge of color space — a neon saturation, a near-black shade — produces percentages that eliminate a lot for some answers and almost nothing for others. Central, balanced colors split the space more evenly, which means every possible answer learns something from them.

The second consideration is naming. The solver speaks in named colors, so guesses you can name exactly beat guesses you can only approximate. "Salmon" is a better tool than "that pinkish-orange I am seeing," and the game itself draws from the same kind of named palette, so the vocabulary transfers both directions.

If you want the fuller strategy — hue families first, brightness later, how to bracket with primaries — the answer page covers it in detail. The solver is the calculator; that page is the method. Together they are considerably better than either one alone.

Intuition drifts. Late at night, on an uncalibrated screen, after three losses, my color judgment is noticeably worse than it thinks it is. The solver applies the same Delta E arithmetic to every row with the same two-hundredths tolerance, every day, regardless of mood. For consistency alone it earns its place.

It also does the tedious part flawlessly: holding every constraint from every previous guess at once. When five percentages are in play, I can hold maybe three of them in my head honestly. The solver holds all five and never guesses a color that contradicts an earlier row, which is precisely the error tired players make on turn five.

What it does not do is feel like a win. Solving by triangulation is satisfying the way filing taxes correctly is satisfying. So here is my honest recommendation, from someone who plays both ways: solve the daily yourself first when you have the time, and use the solver when a streak is on the line or when you want to check whether your instincts were even close. The gap between the two is usually educational.

The candidate pool is a large fixed list of named colors, and the solver pre-computes nothing about your specific puzzle — it filters live from your inputs, so it works on any Colordle-style puzzle that scores with the same percentage system. Old daily puzzles included. The archive and the solver together make a decent practice gym: replay a past day, run the loop, and watch how fast three guesses collapse the space.

The edge case worth knowing: very high percentages. When you score in the high nineties, you are inside a tight cluster of visually adjacent names, and the candidate list can stay stubbornly long because many colors sit almost the same distance from your guess. The escape is a guess from a different part of the palette entirely, even one you know is wrong. A deliberately wrong guess at distance still produces a shell, and its intersection with your near-miss shell is tiny.

The other edge case: rounding. The game shows two decimals; enter them both. 19.7 and 19.69 are different filters, and the stricter you are, the fewer false candidates survive. The solver enforces the tolerance — your job is just to read the number off the screen faithfully.