So pretty good then, eh? But a real test would be for a human to say a statistically significant number of words “at random” and then to measure the distribution of numbers calculated from those. It would also be interesting to do the same thing with words selected by a person from their environment, such as on packages in the fridge.
I’ve read Finnegans Wake, and my guess is the huge number of words drawn from other languages would bias the measurements in some way, but using words from the book would probably still work as a cheap, hands-free dice roller, if you could remember enough of them.
