Steam is also constantly promoting new games to you and making suggestions based on what you currently play. There is also the social aspect in that you can see who is playing what, leading to new discoveries.
I will admit that at least having the verifiable length of time a reviewer played a game helps, but I don't find the reviews themselves to be all that worthwhile for most games.
If it was worth 1500 hours of the reviewer's time then it's probably worth 10-50 of a potential buyer's.
Inversely, I don't need to play a bad game for 1500 hours to decide that it's bad. I played 60 minutes of No Man's Sky and you'd have to pay me to play a single one more. Does that mean I should be discredited from voicing my opinion?
The return system and the achievement system should probably be integrated. Likely there should be some sort of achievement related to 'escaping the tutorial'. They might even call it "Almost bought the farm" or something.
Playing to that point should give a player a good idea what kind of game it is, and what sort of plot (if any) is happening in the game.
From /that/ point they should have maybe 30-60 min of 'game runtime' to return the game or not.
Functionally this would be an in-product demo.
Steam could easily implement this, letting you try purchased games for up to two hours.
It's be trivial for them to implement some anti-abuse system (like "must have purchased N games before using demos, etc).
The exception would be with highly competitive multiplayer games, and even then you can quickly get a very good idea.
Steam isn't bad at discovery, but because they're not creating much original content and their UI is janky, they're also not that good at it.
Source: created comprehensive database of education technology products [1], and while we made best attempt at discovery, we learned pretty quickly it was secondary in importance to having comprehensive info.