Well, it could be another of these threads focused on whining how BG3 cheats dice rolls and it makes the game unfair. Although it is not: I believe writing such complaints is pointless. There always will be diverging opinions, especially when no one did sufficiently thorough statistical research based on good amount of collected data.

But it does not alleviate main problem with this game: it's indeed frustrating for many and frequently might feel unfair - as a subjective experience, not objective truth. Truth does not matter here, only experience. This is a video game, anyway. Its one and only purpose is to give fun to players -- as opposed to attempt at being a precise implementation of some system from the outside of video game domain (whether called D&D 5e or otherwise). It's neither a chess program, a compiler, nor logic simulator for VLSI chips.

Getting straight to the point: my playtime is about 250 hours for now (EA and release). I tried playing both in standard and karmic mode. My general feeling is as such:

Standard dice mode: everyone misses nearly all the time, spell slots mostly get wasted, encounters drag on forever. Prone to streaks and confirmation bias.

Karmic dice mode: much better experience overall (contrary to popular belief!), however enemies nearly always win saving throws. This makes half of the spells useless (high DC and Bane will not help you when they roll base 18 most of the time). Also the description lies -- it does not eliminate streaks, it only skews probability distribution to upper values (you can find a statistical analysis/histogram the net).

Another problem: difficulty in this game is practically equal to zero, as you can save/reload anytime and basically save-scum through everything, nullifying all game's systems. If you have patience to watch these horribly long loading screens, that is.

I was thinking about ways to improve it and a solution came to me which I think is both simple and effective in solving all of these problems once and for all. The idea is to replace (conceptually and algorithmically, but not visually) dice rolling with card drawing, as per following rules:

1. Instead of a d20 you have a deck of cards numbered 1 .. 20, randomly shuffled.
2. Instead of rolling a d20, you draw single card from the top of the deck.
3. When there are no cards left, game prepares another deck. Sometimes a deck can be reset (replaced by a new deck of 20 randomly shuffled cards).
4. There are 5 major decks in the game:
- for players/allies attack rolls,
- for players/allies saving throws,
- for enemies attack rolls,
- for enemies saving throws,
- for ability/skill checks outside combat.
5. In addition to that, there are a couple of smaller decks for damage calculation and other auxiliary purposes, following the same pattern.
6. All combat-related decks are reset when combat begins, i.e. game transitions from non-combat to combat state (first player character enters combat, but not subsequent ones).
7. Ability check-related deck is being reset on short or long rest. It also resets environmental checks (perception, survival, etc.).
8. All decks are SAVED into the game save file.

Consequences of using the system described above are:

1. It IS completely fair and it FEELS completely fair. Both of these properties are mathematically guaranteed. Everything evens up in rather short cycles of 20 draws. So you will always have your share of hits, crits and misses no matter what, and so do your enemies.

2. However, it is still random as you do not know exactly when these hits or misses will come - at least at the beginning/middle of the cycle.

3. Near the end of the cycle you might be able to predict what will likely happen basing on previous results in that cycle. And it's actually cool, because it enables another layer of tactical thinking. Now you think and calculate, instead watching mindlessly how dice roll.

4. Statistical properties are equal to those of unweighted dice. Also there is no room for speculations about dice rigging on game's side. The math is fully open and transparent here.

5. It is impossible to brute-force through the game by simple save-scumming (because of rule 8). On the other hand, it is fully valid to redo a failed fight (rule 6), or even a sufficiently long part of the fight (from cycle start). It is also possible to redo botched ability checks, but for a price of short/long rest. It's up to you to decide whether it is worth to do so (as camp resources are limited).

6. There's still a possibility to use save/reload to cheat a bit -- do some "trial" turns to figure out what the sequence of cards in current deck is, then reload and utilize that information. But again, it requires thinking and planning on your side, instead of stupidly pressing F8 until succeeded.

In my opinion this system would make the game much better. It's really how "karmic dice" SHOULD be implemented, instead of some shady tampering with regular RNG results.

If you agree with me on that matter, please help getting Larian's attention to that proposal. BG3 recently got a high praise and severe overhype, but in my opinion actual experience with this game is inferior compared to other cRPG games like original BG, DA, PoE or even Larian's own Divinity:OS series. And that's a shame, because otherwise the game is excellent. Especially such a great amount of high-quality content is amazing. Some improvement on mechanics side perhaps would make it truly one of the best cRPGs around, as all these overhyped reviews claim it to be.