I compressed it with a simple arithmetic compressor to 4368 bytes (3583 for the guessable words, 785 bytes for the hidden word set), using a zero order model.

I compressed it with a simple arithmetic compressor to 4368 bytes (3583 for the guessable words, 785 bytes for the hidden word set), using a zero order model. I used an arithmetic compressor I wrote long ago at https://github.com/ChrisLomont/ArithmeticCompression https://github.com/ChrisLomont/ArithmeticCompression. Use 26 symbols, and then just feed all the words, a letter at a time, into the compressor. Or, if you want to check it more easily, compute the letter entropy - you get the same result.