I think you'd be best off to go and read a structures book. eg Roark. Torsion of thin walled tubes of arbitrary cross section is non intuitive to say the least. If you want to pile in regardless then the simple formula is derived by twisting a plane section and adding the contribution of each twisted little piece. Needless to say this can fail horribly in pracice once you have re-entrant shapes, or non uniform wall thickness. for thin walled tubes of arbitrary cross section K=4*(enclosed area to mean thickness)^2*t/(length of mean thickness)
For a rectangular tube of constant thickness then K is 2*t*t*a**b*b/(a+b) if t is small
In your research you will come across the membrane analogy. This tells you that the result for a rectangle is the upper limit for a parallelogram of the same dimensions and thickness.
Cheers
Greg Locock