I did some ciphering to try to figure this out. First, some assumptions:
Bellcrank arms are equal length, hence tension on each line due to inertial and aerodynamic forces on the airplane (not counting forces required to react hinge moment) is the same.
Drag and line-inertia components of line tension are ignored.
Handle mass is ignored.
Bellcrank rotational acceleration is ignored.
The first picture shows the things I used in the calculation. I allowed for different line spacing for up and down and for different overhang for the up and down lines. What the heck, why not? I calculated hup from aup, cup, and δhandle, and hdown from adown, cdown, and δhandle.
The second picture is one way to look at the forces on the handle from the lines and the force and moment applied by the flier. After seeing what the guys wrote above, I drew the third picture, which may be equivalent. I separated Tc, the forces needed to react control surface hinge moment, from T, the "common-mode" line tension due to inertial and aerodynamic forces on the airplane. T includes mV2/r and forces from engine offset, the paltrey fuselage lift, and wing lift * sin of the bank angle. An interesting aside is that when Tc = T/2, you've hit the Netzeband Wall.
I attached an Excel file you can fiddle with to see the effect of the parameters in the first picture. The few of you who will want to see the formulas can get them from the Excel cells; they're too hard to type here. Behold that the distance you have to move the point of average finger force, hfingers in the third picture, is exactly the same number as the net moment from common-mode line tension (M in picture 2) per lb. of line tension.
If you set the c distances to zero, as in Sleepy Gomez's clever handles, the only moment (or finger movement) you will feel is from the airplane's control surface hinge moment. "Overhang" adds some artificial feel. Why would you want that, and how much would you want? My guess is that you want to get some feedback at the handle proportional to how much you are actually deflecting the control surfaces. Because the amount you can deflect the control surfaces (which are on the far end of a 70-foot spring) is a function of the common-mode line tension T, you might want to get some of this feedback from common-mode line tension via overhang. I don't know offhand how to calculate how much overhang that is, nor how I'd know if I got it right experimentally, nor if my hunch about what overhang is for is correct.
Edited to fix Excel file. Tangent, Arctangent, what's the difference?