I believe, that some comments are appropriate:
* For two ideal complementary transistors the transition from high (VDD) to low output level occurs at a common input voltage of VDD/2. The midpoint of this "quasi-linear" transition region also is at VDD/2. Therefore, a feedback resistor Rf between output and common input creates the correct bias (which is dc stable because of the negative feedback action).
* However, when the common input node is connected (via a capacitor) to a signal voltage the feedback resistor Rf also acts as a load resistor. Since the MOS pair acts as a current source, the load resistor determines the voltage gain of the whole circuit. Therefore, with an Rf value in the order of - let's say - 100 kohms you easily can get a gain of 100...500.
* Since this gain value is strongly dependent on MOS parameters it makes sense to apply signal feedback. For this purpose the input signal is applied through a series Rin-Cin combination. Then, the resulting gain is app. Rf/Rin. For this combination, stable gain values of 10..20 are not uncommon.
Added later: The CMOS pair acts (in principle) as a current source because resp. if the internal MOS output resistance is neglected.