The software they use will calculate the 3D location of this body part to generate the attacking players (onside /offside) line.
It seems not.
https://twitter.com/DaleJohnsonESPN/status/1191045286589878272When the body part in question (shoulder) isn't grounded a CoG (centre of gravity) is marked manually using the feet. The software then creates a vertical line from the shoulder and crosses it with a perpendicular one from the CoG to determine it's distance from the goal-line.
I've watched a few videos but have been unable to find a proper explanation of the compete protocol for determining the CoG (which foot, heel or toe etc.) and in any case using a 2D image is never going to be exact.
The absurdity is best explained by what appears to have happened in the Villa game (working backwards):
Firmino's right shoulder has been treated as if it was directly above his left foot (his right foot won't be 100% perpendicular with his right shoulder but it would be far closer).
The VAR has adjudged that Firmino's boot (which could only have been his left) can be seen between Ming's legs after previously having to estimate the CoG based on the belief that his feet were obscured (which is why they re-drew the lines). The role of the replay operator (not a trained referee) is unclear in determining the CoG.
However, based on 1:18 of this video played at 0.25 speed (
https://www.youtube.com/watch?v=YdFKNBzQ_Ic) Firmino's left foot is clearly off the ground which is important because the software treats the CoG as grounded.
In summary the CoG is a guess which renders the entire process useless for close decisions.
ps. We didn't get to see how Ming's CoG was established.