Substitution cipher decoder

With a key: type the 26-letter key, the cipher letters for A to Z in order, and encode or decode above. Without a key: paste the ciphertext into the solver and press Solve. The sample is the opening of Moby-Dick, encrypted with a key the page doesn't know.

The solver works best on 100 letters or more. On a short message a few rare letters may come out wrong; fix those by hand, the way you'd finish a crossword.

It guesses a key at random and scores the decryption by how English it looks: every run of four letters is checked against counts taken from three public-domain novels, so THER and TION score well and QXZK scores badly. Then it swaps two letters of the key and keeps the swap if the score goes up, over and over until nothing helps, and starts again from fresh guesses, keeping the best. It's the same frequency analysis people do by hand, done thousands of times a second.

The statistics are downloaded the first time you press Solve, as one file. Your ciphertext isn't sent anywhere.

Count the letters: in English, E, T, A, O, I and N are the most common. Look at one-letter words, which are almost always A or I, and three-letter words, where THE and AND are favourites. A repeated pattern like XYZZY might be a word with a double letter. Each letter you place makes the next one easier. The 9th-century scholar al-Kindi wrote the first known description of this method.