You *could* calculate it from the text itself, but going that route pretty much means you're building your own RTF parser. There are a lot of keywords to pay attention to that can force or delay page breaks, such as tables, footers (destination text), images, individual page margins, varying fonts, kerning, yadda yadda yadda.

I would guesstimate: you know from the spec you linked that a page defaults to 792 x 612 pts with 72 pt margins on the top and bottom and 90 pt margins on the sides. And you know the font size in points, so you know how many lines can fit on a page. Figure out the length() of the text and guess. It won't be precise, but it may be good enough.