In an English text document with fixed 92 character line breaks, what is the probability that at least one line will be exactly 92 characters long?
The document is 125 lines.
I want to determine how reliably I can determine column width by examining the existing document contents.
I have no idea where to start.
You need the distribution of the number of letters in a word in English where a word that has a trailing punctuation mark adds one letter. For example, if every word were $5$ letters long except for the last word of a sentence words would always end at $5 \bmod 6$ letters and you would never have a line come out even. Given the distribution of word lengths you could then figure out the chance a line comes out even. You could do even better if you knew the distribution of pairs of words. I suspect that if you have a one letter word the next word is more rarely two letters (although I am is reasonably common) than normal.