Frieder mentioned to use a sample rate which is multiple of 50Hz, means 100Hz for instance.
You are totally right, suppressing of 50Hz hum by the use of 50Hz sample rate is impossible! Assume for instance that you take a sample at the maximum of 50Hz hum sine, then due to a choosen smaple rate of 50Hz, next sample will again be maximum voltage of 50Hz hum sine. And all the other samples too!
Totally different performance when sampling 100Hz though, two succeeding samples will always vanish, because they show same level but of opposite sign. 50Hz is optimally suppressed, provided number of samples taken into consideration with averaging is even.

Unfortunately, hum mostly contains not only 50Hz sine, but also lots of harmonics, especially 100Hz, when a rectification takes place somewhere. Then, sampling at 100Hz, can make the situation even more troublesome. Yes, here choosing a sample rate which is a multiple of 50Hz is not at all a good idea! Then, a sample rate is to be choosen, which's harmonics must not coincide with the with harmonics of hum. So, a sample rate of about 60Hz is indeed a good choice, first coincidation will be at 300Hz.

Because it's mostly the harmonics of 50Hz hum which can make trouble with the averaging methode, the only remedy in most cases is to choose a proper anti-aliasing filter at input of sampler. So, a high order low pass filter (Bessel) with a corner frequency of let me say 10...20Hz in combination with a sample rate of 100Hz would allow a perfect suppression of 50Hz hum and all higher harmonics.

Kai