OK, so taking the model of m mass/unit length, length C on the X axis, length A above the X axis by angle α, and length B below the X axis by angle β we get:
m(C/2)*C = m(A/2)Acos(α) + m(B/2)Bcos(β)
and
m(A/2)Asin(α) = m(B/2)Bsin(β)
All the ms and 2s drop out and you get:
C^2 = A^2cos(α) + B^2cos(β)
and
A^2sin(α) = B^2sin(β)
OK, I move the A^2* cos(α) to the other side, square both sides of each equation, then add the equations together to get:
A^4 + C^4 – 2(AC)^2*cos(α) = B^4
Or
α = acos((A^4 + C^4 - B^4)/(2(AC)^2))
if you move the B^2*cos(β) over instead and repeat, you get:
β = acos((B^4 + C^4 - A^4)/(2(BC)^2))
or, just what MikeS said.
Square the Pythagorean theorem and substitute for the C term in the numerator of the alpha equation. This gives:
cos(α) = (2A^4 + 2(AB)^2)/(2(AC)^2)
or
cos(α) = (A^2 + B^2)/(C^2)
Using the Pythagorean theorem again that gives us 1/1 or α is zero. Same with the β equation.
So, if you have a Pythagorean triple, you need to the two smaller lengths together exactly opposite the big length.
ETA: Naita discovered that this holds for point masses when the “length and mass are the same” which is a bit specious, but the math then works out the same for distributed mass lengths.