> is there a potential for selection bias?
Yes, but it beats blanket statements like "D&D games are BY FAR more popular in gaming history".
Let's compare two very similar games: BG1 and Fallout 2. Came out at the same time, have about the same amount of (modern) sales, both have great story, atmosphere, dialogues, and such, both have turn-based tactical combat. The non-D&D game has 50% higher positive feedback.
Other games are harder to directly compare.
Mutant Year Zero and Pathfinder of the same release year. MYZ has great story, worldbuilding, dialogues, adventure in general, and it's got turn-based tactical combat with action points, but it's of lesser scope than Pathfinder. Still, MYZ has 270% higher positive feedback. It's also 60% less popular, but what else am I going to compared it to? DOS2? DOS2 annihilates everything. Gears Tactics? That's a really niche linear W40K game, although of high quality. XCOM 2? While they're at the same approval level, XCOM 2 is ways more popular, and considering that XCOM 2 is basically nothing but combat, it shows what kind of combat system is more popular with the gamers.
As for BG3 e.a., there's absolutely nothing comparable, at least to my knowledge, not in 2020, not now. Maybe Dragon Age 4 would be a good comparison? But it stopped being tactical turn-based after DA:O, so that's a negative. Maybe, if by some miracle it's turn-based, we'll be blessed by some interesting numbers.
(Edit: probably not though, EA got off Steam so no stats for us)
> I imagine classic RPGs and older games were making the bulk of their sales and positive reviews prior to being placed on Steam, and that most people who owned the game prior to Steam are not buying the game again.
You can compare modern Steam re-releases to each other, like Fallout 2 and BG. I couldn't find reliable sales numbers neither for original Fallout 2 nor for BG, sorry.
