I think you could get away with using a double zeta basis. Some studies have actually shown cc-pVDZ to give results more consistent with empirical observations than any of the higher zeta sets. Triple zeta would be fine, though, to answer your question. If you are looking for a minimum (not a TS), you might try using MP2 instead of B3LYP.

I think its good. For now I could see in my opt that TZV gives energies lower than 6-311G. In other way, you have three times as many functions as the minimum basis, so the interaction with each carbon have a better description. Other possibility is DZV that have a very good description for organic compounds.

Other think that could be, in general good, is use a CBS method. If your molecule it isn't too large, could give you a lot of information, including thermochemistry energies with an excellent accuracy.

If you use a gamess, you could use a G3MP2 method because CBS are not currently active.

Since selenium is not a too heavy element one can use at least 6-31g(d) basis states (d refers to polarization functions). I also suggest Gaussian 09 program because it contains many kind of basis sets and methods.