Most efficient way to do front-back test

Your method is the correct way of doing it, but you don't need to square root anything, because you are only interested in the sign of the dot product's result. So the vectors don't have to be of unit length. So if your front vector is V1, and the vector from the target to the main character is V2, you just have:

dot(V1, V2) > 0 -> the target is within a 180 degree curve of the main character in the direction he's facing
dot(V1, V2) < 0 -> the target is "behind" the main character
dot(V1, V2) = 0 -> this one is a special case where the target is "in the corner of eye" of the main character, so how you define it is up to convention (it doesn't matter)

This calculation only requires a vector subtraction, a dot product (= 2-3 scalar multiplications and 1-2 scalar additions depending on which dimension you are using) and a scalar comparison, per character to test.

oh yea. the projection of one vector onto another should be good enough. Lol, y didnt I think of that. Thats y we have the forums i guess.
Thank u sir Bacterius.