How do you define a game being universally good? Is it a technically advanced game? A game with lots of depth? A game some people love a whole lot?
I'd say the best way of defining the "goodness" (quality, if you like) of a game, is to say how many people could potentially like the game, and how much they would like it. Essentially, the bigger the potential love for the game, the better the game.
I don't see any other good way of defining the goodness/quality of a game. Now, the main problem with this definition is that it's ridiculously hard to measure. Measuring the potential love is currently far beyond our capabilities (I hate calling things impossible). What we can do, though, is make an assumption of how loved the game has been by the people who bought it.
We can do that by either looking at two things. How many people played the game, and how much they like it. How much people like the game is hugely subjective and quite hard to measure fairly, but how many people played the game isn't.
If people paid to be able to play the game, they most likely thought they would like it. As they thought they would like it, they generally did.
Essentially, it's like this. We know how many people bought the game. The vast majority of people who bought the game liked it. From that, we can say that a game that sold a lot is a great game, because it's much loved. We don't know how much loved it could potentially be, as we have no way of measuring that (yet). As we don't know how much fun a game that sold little could have brought, we can't say that it's a poor game. We just have no clue how good it really is.
Or in simple terms
A game not selling much doesn't mean it isn't a great game.
A game selling a lot means it's almost always a great game.
Derivating quality from sales isn't a brilliant way of doing, but it's better than all the other methods we have.