If you think of reviews not as score but subjective opinions and do not try to fool yourself by trying to come up with some magic function that extracts all the valuable data from them and nullify subjective judgement, they are valuable. The hard task is to read them and extract valuable information, so that sometimes the effort does not worth the result, but this one is also a bonus, since non-structured data is hard to game.
> If you think of reviews not as score but subjective opinions
I do think of them that way. But I have no context in which to interpret those opinions, so they are valueless to me. If I don't know what the reviewer values and doesn't value, I can't know how to interpret the review.
It's like with movie reviews -- a movie review from someone I don't know is without value, because I don't know what their tastes are. Did they dislike the movie because it's bad, or because they don't like that particular style? There's no way of knowing.
This all would be more workable if all reviews were honest, but they aren't. When you toss in the need to try to discern what reviews are real and what reviews aren't, the entire task becomes impossible.
This is why I've learned to ignore all reviews that aren't from someone that I am already familiar with.
