king_kong

You could calculate it from any frequency, but the trick is when taking it from your frequency to an annual basis (which I think is how most people would compare things to one another).

If returns are normally distributed and there's no serial correlation, you can multiple the sharpe ratio by (f_end/f_begin)^0.5 to get an approximation of the annualized value, where f_end is the end frequency and f_begin is the beginning frequency (normally 1 initially).