BG3 is definitely a sweaty DM that most real people would not want to play with in a real situation, but not really for the reasons you're focused on here - these are their own problem though, don't worry!
You probably know this, but just for clarity and background: Video games don't have the ability to actually roll random dice - they use maths; specifically, algorithms with a variable amount of chaos-sourcing inputs, in order to generate a string of numbers that can seem to mimic, or simulate, actual random rolls.
A lot of players put in a lot of time and effort to gather rolling data during the EA, and to assess the game's RNG, and what was found back then was that Larian's RNG was not very good; it was lacking in adequate sources of chaos, and produced a noticeable wave pattern indicative of the algorithm it was using to generate its numbers. Many people claimed to have no problem, or noted the prevalence of confirmation bias in individual cases, but most of the people who took substantial samples and seriously looked at it found the results to be unsatisfactory.
Larian's response was to introduce 'loaded dice' - which is basically the game putting its fingers on the scale to fudge the results, all the while using language that functionally calls the player a cheater for using it - it's turned ON by default, and you have to turn it off in your settings before you play.
The intent of the loaded dice was to break up failure streaks and prevent players from being crit too often, and so on... the result, however, was that it fudged the dice for everyone, including enemies. Just as an example: If you have a particularly high Ac, and most enemies miss you most of the time (or should do), you'll frequently see enemies rolling in the 16-19 bracket, just to hit you, so that it can 'break up' that streak of repeat misses from the enemy: high AC = you get hit MORE, because the enemy rolls better against you. They'll also crit you more often. That's just one example of ho it ends up working in practice. The same theoretically applies to you, the player, but there are always far more enemies than players, so you will almost always be on the receiving end of the loaded dice fudge.
So, the first question is, were loaded dice turned on or off when you did your tests?
I don't believe anyone has tested the averages of damage rolls specifically, in our past collations; we were mostly focused on the d20 rolls, so that might be interesting data to see.
