For years, somewhere between 41 and 44 grains of IMR4064 tested under 1/2 MOA through 300 yards in a given quality barrel in a well built rifle properly tested. In twists from 1:10 to 1:12. Decent barrels will shoot sub MOA through 300 yards with a 5% spread in charge weights across different lots of ammo. Many matches won and records set with it.
IMR4895 was and still is a close second, in my opinion. But it shoots 150's and 155 more accurate through 1:12 and 1:13 twist barrels than IMR4895; so does Varget.
Varget may well duplicate IMR4064's performance with 168's; never shot it my self.
1:12 twist barrels 24" and longer will well stabilize 190's leaving at about 2500 to 2550 fps from .308 Win cases. Shot many of them atop 44 grains of IMR4320 in bolt guns and Garands. Won my share of matches doing so.
That said, if you're judging accuracy on one or two few-shot groups, any load may well shoot a tiny one. If you want a lot of confidence your load's a good one, shoot at least 20 shots per group with each one, then pick the load that shot the smallest one. As there's a 5X or more spread across several few-shot groups and the smallest one's never the first one shot, such tests are not very representative of the ammo's performance for accuracy. Velocity, too, for that matter.
And ball powders are not used for best results by record setters and match winners producing the best results for accuracy.