Yes, I spent some time playing with the weak assessment engine they have here and came to the conclusion that g3 was good for white typically in lines where a6 is played. I play d4 and before that it was Nf3 and then c4 earlier still. This is the first time I've ever looked at Scandinavian lines. I think that a lot of 1. e4 players wouldn't consider g3 lines. Even some stronger ones, more than likely. I have the same reaction to 0.9 from SF. It's inflated, normally. It usually means about 0.5, after a few optimal moves.
Oh i know what you mean about eval inflation. Earlier engines like stockfish 11 gave far more conservative estimate despite having usually similar evals if a position is slightly better or much better etc at higher depths.
But the problem here is not one of eval numbers. The problem is if you analyze some of these lines deep enough, black is like one inaccuracy away from being completely +2. This isnt some old indian position where the engine may give an inflated but steady 1.2. This qd6 a6 scandi stuff looks like a total minefield for black whereas if white misplays it , his advantage just becomes merely a standard white first edge in a secondary defense. That is alarming. It means black faces great practical difficulties which are even more worrisome than the theoretical eval.
if you look at some of the database stats, they dont paint a pretty picture and if you look at gm games, even when white botches it with an inaccuracy his positions are far more pleasant and less liable to decisive errors. Whereas black cannot trust natural moves for quite some time.
After 4. d4, black's best seems to be ....a6, which means black can meet Ne5 with Nc6. Then, white can't follow up with f4, so Shirov's line is completely wrong.
You obviously dont play the shirov line if black abandons his standard piece formation, but after 4...a6 5.g3 white's advantage is still quite large, and more importantly, idk for what. white has more space, nagging pressure on the light diagonal, will harrass the queen some more with bf4 etc. All black has, is a funky position that looks an inferior cousin of the d5 nimzowitsch as blacks queen is harrassable on d6 and he really cant hope .to play e5 in the near future.
for example. 4...a6 5.g3 bg4 (main move here) 6. h3! bh5 7.bg2 nc6 8.d5! nb5 9.bf4 qc5 10.be3 qd6 11.qe2 is just busted for black. I welcome you to suggest improvements for black
the most i can say in defense of the black side of this is that the number of master level players that would know about some of these lines right out of the top of their head is quite small. even those that do to some degree often dont execute it perfectly and +1 advantages due to inaccuracies become more manageable white edges. But in an age of wide access databases and cloud engine supercomputer in every player's laptop why would you risk getting into a miserable position like this!? it doesnt even give you that many practical chances?
I can tolerate a 0.9 from stockfish 14 if the position looks sufficiently complex for humans to be worth the dubious eval and i wont be limping for a draw 15 moves later, but this just looks like black merely wants to prove he can survive this lol