This is cool. The given example of typos on page kind of highlights the fact that more sophisticated math involving a prior might give better results in some c

This is cool. The given example of typos on page kind of highlights the fact that more sophisticated math involving a prior might give better results in some cases. The beta(1, N+1) prior is an assumption that you start with the a priori knowledge that a typo rate of 1% and a typo rate of 99% are equally as likely as each other. Most people would assume that books don't have typos on most pages and a 99% typo rate is unlikely. However as your sample gets bigger the prior matters less and less so this rule is still useful. Just know that it is reasonable to bias the results a bit according to your prior when N is small.