It is an observable fact that there are two fixed points on the sky around which all the stars appear to rotate. These points (the celestial poles) are located 180 degrees apart, creating the illusion that the sky is a huge sphere surrounding the Earth.
Polaris is located 40' from the actual point of rotation in the north and so traces out a circle on the sky which is 80' across or just over 2 Moon diameters. Likewise Sigma Octantis, the brightest star nearest to the south celestial pole is located at a distance of 105' or just under 2d from the point of rotation in the south.
In the north the stars rotate around the NCP anticlockwise while in the south they rotate clockwise. If you visualise the Earth as a sphere with an axis that connects points on the surface that coincide with the NP and SP such that they point almost directly at Polaris and Sigma Octantis respectively this can account for what we observe directly.
How can this be replicated if you model the Earth as flat. In which case there can be no point of rotation in the south because the south 'pole' becomes the circumference of a circle and not a point. That means the south celestial 'pole' would also be a large circle and not a point of rotation.