So I'll come clean. The catalyst for me to be thinking about this subject, is horse racing. I've long since been perplexed as to why the outcome of photo finishes is difficult to predict. Indeed the challenges of identifying which horse has won is a function of the broadcast camera being out of line with the focal plane of the winning line
VAR faces the exact same challenge. If the cameras merely needed to identify the location of the feet on the two dimensions of ground level, the process would be quite straight forward. However, as soon as the location of any other body part is required, it actually becomes hideously difficult.
VAR's attempts to manually draw vertical lines is very crude and flawed IMHO. Indeed, inaccurate attempt to marry virtual offside lines on the vertical axis with an offside line on the ground lacks the precision of the single pixel (now 3 pixel) lines that the final decision relies upon. Using imprecise and very precise methodology simultaneously to determine exactness is absurd. And, one of my major gripes with VAR and offside is the time taken to reach a decision, almost all of which is consumed by the crude vertical drawing of lines
So whilst I think what they've attempted to do hereto has been moronic, the World Cup trial will probably represent a major improvement because the technology promises to identify offside position on a fully automated basis (possibly within half a second)
I personally predict the AI will develop super fast and surpass some people expectations, but that's another discussion
Thing about photo finishes, is you can bet on the outcome in real-time whilst the Judge determines the winner. So I have a vested interest in the subject