Choosing Single Base can be better than choosing a Virtual Base Station, but only under very specific conditions. 

A Single base solution degrades with distance. If you have a single base (that you can see from the rover or under a km) or a station within a few km, that may yield better results than VRS on days with good ionospheric conditions, but that may vary from day to day. One rule of thumb is to not use single base if you are more than 10km from a station or base. Another is to consider using static methods if possible.

For single base to work its best, the base has to be set up in as good, or better, sky conditions than the rover. That is, with low multipath conditions, and it has to support the same constellations as the rover; and you have to be sure it does not get stolen.

It is possible (but not recommended) to get fixed results from single base at very long baseline lengths, 20km, 30km, and more. The problem is that you will not get consistent results at long baseline lengths as the conditions could vary from day to day, even hour to hour.

Contrary to some marketing-driven rumors from the early days of RTN, VRS does include a physical base station (PBS or PRS); a PBS code is in all VRS corrections. The model of the corrections is virtual, but there is still a tie to a fixed station. Nearly every rover that can export vectors can export the rover-PBS vector.