[Evilhomer]

The inaccuracy of indypendents example is that the unknown "head" probability has four positions to be at (so you would have to compensate with a factor of 4 here), while each tail in the example only has one position (while having four "cases" instead).

Probably pretty fuzzy explanation, but if we simplify to Baalz original 4 random coin flips, 3 known the chance for the last unknown is 50% heads and 50% tails (as Baalz wrote).

That said I am not sure about the mechanic behind site distribution so I am not sure how well the coin flipp theory works. It is however always better to search provinces with few sites for the simple reason that 4 sites is the maximum (thus you cannot find 2 new sites, if you have already found 3 for example).