In fact, upon closer inspection, it does not seem too hard to determine ##g## starting from a large angle either. If you have a look at that Wikipedia-link I gave and you go to the section "Arbitrary-amplitude period", you can see that ##T## is the product of ##4\sqrt{\tfrac{\ell}{g}}## and an elliptic integral that depends on ##\theta_0## (the amplitude), but not on ##g##. So, if in your experiment you measure ##\theta_0## and ##\ell## and then compute the elliptic integral numerically (or from a table) using your measured value of ##\theta_0##, you can determine ##g## this way.