The curious thing is it was about my original idea, but it was turned down by my supervisor. Not without reasons too.
First it is meant to be able to be free to rotate, part of the task is bringing the pendulum to top position from hanging down. But I could find parts like that I think. The other problem is friction. Most encoders are practically friction-less. I wouldn't like to see my model remain in "unstable ballance" position after being switched off. And then precision - temperature drift of resistance and such could ruin the results by shifting the "zero position".