"John Martin" wrote: (clip)Since those forces are acting on diagonally
opposite corners of the
nut, can't you look at the forces on each corner as two vectors - one
perpendicular to the face of the nut (and the face of the wrench), and
one at right angles to it? (clip)
^^^^^^^^^^^^^^^^^^^^
Yes, the forces can be resolved into two vectors, as you say, but if the
wrench is properly adjusted, the forces are essentially at right angles to
the jaw faces, so the second vector component is negligible. What makes the
difference is that one of the forces is close to the base of the jaw, and
the other is nearer the tip. Since the forces are equal, the one nearer the
tip creates more bending moment. You want this to be in the stronger jaw.

The movable jaw transmits its moment into the slot in the handle, which is
likely to be weaker than the stationary jaw. So you want it to receive less
moment. Therefore, turn the wrench so the force on the movable jaw is away
from, the tip.