The main problem in making the game more accurate in terms of when to declare an advantage or not will be the percentage of success lowering down dramatically at low level. Will players like the fact they have to wait 5 rounds to kill just one goblin? I am not sure honestly.
I guess they began implementing stricto sensu 5e combat rules but realized it would be too punishing to the pace of combat. So what's the best compromise here?
Those rules are such a sensitive small ecosystem, aren't they? Inspired by this reddit post I made my own analysis. I've ran a simulator with 1 million of goblins and let them fight with one immortal 1st level Fighter (STR 16).
Current BG3, Goblin AC8, HP14, 20% of attacks have disadvantage or they are normal, 80% have advantage
Code
Rounds Avg Min Low Med High Max StDev
Short Sword 2703117 2.7 2 3 3 3 8 0.6
Warhammer 2425818 2.4 2 2 2 2 8 0.6
Greatsword 2018406 2.0 1 2 2 2 7 0.4
RAW DnD, Goblin AC15, HP7, 20% of attacks have disadvantage or advantage, 80% are normal attacks
Code
Rounds Avg Min Low Med High Max StDev
Short Sword 2723422 2.7 1 2 2 2 21 1.7
Warhammer 2498914 2.5 1 2 2 2 23 1.7
Greatsword 1972064 2.0 1 1 1 1 19 1.4
We can clearly see that Larian tries and actually succeed in achiving roughly same total combat durations. Average time to kill a Goblin is same in BG3 as in RAW DnD. Looking to minimal/maximal value and medians (low, middle and high) there is one round difference, which is OK by my standards. The Problems of RAW are very long miss streaks and larger standard deviation which leads to bad feeling and discomfort. So I think if we are trying to persuade Larian to revert all their changes regarding advantage/disadvantage and HP/AC back do RAW DnD, we should suggest an alternative system which deals somehow with those miss streaks.
My favorite solution is simply not interpret all misses as misses. Often you hit but deal no damage. There could be an animation, sparks, taunts, shouts, etc, depending on how close you were on your d20 attack roll.
Other solid solution would be pseudo-random distribution like Dota 2 is using. This will fake die rolls depending on you current streak but maintain the average probability the same.
