26-06-2026, 05:38 PM
About conditional character entropy (h2), you can make this arbitrarily high by allowing for a huge glyph set. An easy way to do this is by looking for tiny differences between versions of your chosen glyph (like "a" or "o") and assigning a unique character to each one. There are other ways as well.
Well, to be more precise, the absolute number of different glyphs (h0) is not relevant, but h1 is. For example, if I use one special character "ç" in a novel, this will increase h0, but it will not affect any other statistic. However, if I randomly replace half of "c" by "ç" in a text, this will increase h0, h1 and h2.
Put differently, if you create more chaos by adding new glyphs in large numbers, you do increase conditional entropy but also the size of the alphabet and the way it is balanced. Therefore, h1 and h2 are best compared in tandem.
But when I talk about the entropy problem, that's just shorthand for saying "q is always followed by o, why do so many words end with "n", why is this "n" so often preceded by a number of minims, which in turn are preceded by "a", what's up with "ed, ched and edy", what's up with gallow placement, and if we ignore the frequent suffixes, what's actually left of the root of words? etc etc etc". There are very concrete problems lurking behind these statistics.
Solvers ignore this and keep ignoring this when they have it pointed out to them, because they cherry pick words that "work" and then creatively interpret the output. They all choose their own way of adding information to the text that isn't there (= increasing entropy), which is why all their solutions work just as well.
No solver understands that the script is the most normal part.
Well, to be more precise, the absolute number of different glyphs (h0) is not relevant, but h1 is. For example, if I use one special character "ç" in a novel, this will increase h0, but it will not affect any other statistic. However, if I randomly replace half of "c" by "ç" in a text, this will increase h0, h1 and h2.
Put differently, if you create more chaos by adding new glyphs in large numbers, you do increase conditional entropy but also the size of the alphabet and the way it is balanced. Therefore, h1 and h2 are best compared in tandem.
But when I talk about the entropy problem, that's just shorthand for saying "q is always followed by o, why do so many words end with "n", why is this "n" so often preceded by a number of minims, which in turn are preceded by "a", what's up with "ed, ched and edy", what's up with gallow placement, and if we ignore the frequent suffixes, what's actually left of the root of words? etc etc etc". There are very concrete problems lurking behind these statistics.
Solvers ignore this and keep ignoring this when they have it pointed out to them, because they cherry pick words that "work" and then creatively interpret the output. They all choose their own way of adding information to the text that isn't there (= increasing entropy), which is why all their solutions work just as well.
No solver understands that the script is the most normal part.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)