The paradox is given in the chap. 1 of the book titled : Mathematical Fallacies and Paradoxes, by Bryan Bunch; as given here.
The book explanation has no mathematical formulation, say if states the path in terms of cycloid (also shown in Fig. 1.5, on page #7); then there is no formula for the same.
Also, using that the nearly flattened ends of the curve for inner circle's point B could have been given a mathematical explanation.
Further, there should be some mathematical explanation & quantification for the reverse loop attained by the fastest point on the rim, as shown in Fig. 1.7 on page #9, as also shown below.
Also, I am not clear if the reverse path is restricted to only the fastest point, or decreases quantitatively with each inner point till the center.

There are two motions involved.
One is the horizontal movement, another one is the clockwise rotation.
Let the initial position of the center of rotation be $(0, r)$, at time $t$, its location is $(vt, r)$. This point is not rotated.
Let $\gamma$ be a real number that will indicate the distance from the center for points that are initially at the same vertical position as the axis and let $\omega$ be the angular velocity. The locus can be written down as
$$(vt-\gamma r \sin (\omega t-\theta), r - \gamma r \cos (\omega t-\theta))$$
If we differentiate the first coordinate, we get
$$\frac{dx}{dt}=v-\gamma r \omega\cos (\omega t-\theta)=\omega \left(\frac{v}{\omega}-\gamma r \cos (\omega t-\theta)\right)=\omega y$$
Hence $\frac{dx}{dt}<0$ if and only if $y<0$.
That is any point that can go below the track will experience a negative velocity in its trajectory.