Not real-life effect on accuracy (which has so many variables in it that it is a near impossible task anyway), but Bryan Litz does this using results in MV SD values for six cartridges (223 up to 338LM size/volume wise) with seven powders, six of them extruded and one (Magpro) a ball type. Each cartridge was tested and chronographed twice and some three times (these having undergone a primer substitution) with charges obtained from a powder measure, RCBS Chargemaster, Satorious lab scale.
These tests can be found on pages 117 - 133 of Modern Advancements in Long Range Shooting Vol II by Mr Litz and available from Applied Ballistics or its approved outlets.
With the exception of .300 Win Mag which had both extruded (H1000) and ball (Magpro) variants, every cartridge saw large SD reductions in the move from thrown to Chargemaster dispensed, but with smaller reductions going onto the Satorius weighed and adjusted charges (to plus or minus 0.02gn), although the improvements reduce as the cartridge and charge size increase. For the 300 Win Mag, the pattern applied to H1000, but Chargemaster SDs were poorer than metered examples with Magpro. (This doesn't surprise me - ball powders flow too freely for the Chargemaster IMO and I won't use it for this type.)
Other factors to consider are cartridge / charge size and weight as a plus or minus 0.1gn variance has a great deal greater effect on pressures and MVs in a small cartridge than a large one. Then there is group size sensitivity to charge weight variance. In an ideal world all our loads would produce one-hole groups across 0.5gn charge spreads, but unfortunately over the years, I've had to live with many where a 0.2gn change doubles 100 yard group size. Finally, if the rifle and cartridge group well at all ranges, how much elevation on the target is injected by MV changes - that obviously being mostly a function of sheer distance.
When I shot a lot of 223 Rem with 90s in FTR at up to 1,000 yards, really consistent charges were vital to long-range performance so everything was weighed and adjusted to plus/minus 0.02gn on a set of Acculab VIC123 scales. That's plus or minus a single kernel of Re15 which was what I used at the time. Loading for this rifle was a slow tortuous business! But for a lot of cartridges, being shot over short to mid ranges, the ChargeMaster is consistent enough, provided that consistency is regularly checked throughout the loading process. Many times I've seen the device behave admirably, often producing less variation than the spec, over quite long strings of charges being dispensed then for no apparent reason, it starts to play up or move a full grain away from the setting. Sometimes, recalibrating it works, other times it needs to be switched off and 'rested' for an hour or two.
Incidentally, mentioning Modern Advancements in Long Range Shooting Vol II this is only one of several such exercises covered in the book - bullet pointing and trimming; effect of neck tension changes / annealing; flash-hole deburring and chamfering. It also has some very interesting stuff on measuring groups and a very long range-tested report one of those perennial range-house discussion and forum topics that has no doubt ruined a few friendships over the years - 'angular convergence', otherwise known as bullets 'going to sleep' and shooting smaller groups in MOA terms at 300 and longer distances than at 100. There are tests of 22LR ammunition and Laser Rangefinders and a quite fascinating section on barrels including two types of carbon fibre wrapped models and a new type I'd never heard of, the Teludine Tech StraightJacket Barrel System or SJBS but which seems to work.
