I don't think there is a way to anticipate the possible effect of the presence of the His-tag on the protein behavior. Therefore the rule is to try with the tag at one end and see what happens! If there is something wrong, have a go with the tag at the other end. And if it is not working properly with this tag, try an other one...
I agree with Dominique: which position will give the best expression of soluble product depends on each protein and is impossible to predict. However, my preference as a first choice would be N-terminal, because most expression vectors include a recognition sequence for a specific protease between the tag and the coding sequence of the protein to be produced. This enables removal of the tag while leaving only one additional amino acid (or possibly zero in the case of enterokinase) fused to the genuine sequence of the desired protein. This wouldn't be the case for C-terminal tags, because most proteases used for tag removal cleave between the penultimate and the last C-terminal residue of the recognition site. Thus, in the case of a C-teminal tag, removal of the tag would leave most of the recognition site sequence fused to the sequence of the recombinant polypeptide.