Testing depend largely on good Openings

[Marcus91]

That is exactly what i try to do - create map or collection of dubious opening lines. Currently i covered 2 ply and 3 ply openings.
Not so much in 2 ply: just 1. a4 f6 and 1. h3 h5.
At 3-ply covered and approved 86 lines.
4-ply in process (674 lines in consideration to be good, 4140 lines remains to check).
Tree expands numerously be each next ply.

I think most of that information is uninteresting because one of the sides will never play one of the moves, so those positions will never happen.

It's more interesting to have one side playing the strongest moves, and the other side playing the dubious moves (say, we have a strongest attack that beats 1.g4, but it only requires a winning move by black, and we prune the rest.)

This way the tree will not expand as much and we'll know where the draw line is.

From point of game end result there is no difference you play 1. e4 or 1. d4 or 1. g3 or 1. Nf3 or 1. f3 or 1. h3 - objective it is draw anyway.
There isn't so much theory for such moves as 1. h3 that is reason why humans and GM do not play this move.

stating whether some moves are good or bad based solely on the outcome of a perfect game is not a good way to choose moves. The best move must also be based on which one gives you the best chance of winning assuming an imperfect play on both sides.

I am already awaiting this argument.
Yes, moves has the different winning chances, usually you must play moves which maximize your chances.
But now look this way ...
You know your opponent is well prepared against your first move 1.e4.
1. g3 or even 1. h3 gives you " ... best chance of winning assuming an imperfect play"?

Probably not at top level, it might be good to play it in blitz games, actually some doubts opening are played in short TC