Oh yes, let's not forget theory three: *Wow* sold well because it's a really good game. Coming in at 93% on gamerankings, there's even something of a semi-objective measure there...it's the most highly rated MMO of all time.
I'd love to believe that's all there is to it...but articles like this leave me little hope. And then there are games like Ico: high ratings, poor sales...and then there's Spider-Man 1, considered marginally above average by the critics, one of the top ten sellers of all time. Mark Nau once ran a correlation between gameranking scores and sales, and although there is a correlation (yay!), it's terribly faint.
Still, quality helps, and if you're the best in your class, it probably helps a lot.
One trap I'm falling into is the "what's so special about X" trap: frequently, when a game is succesful, we try to analzye it, and look for the one factor that made it such a success. When Myst was succesful, it was "This kind of graphic adventure must be hot!" When Tomb Raider was succesful, it was "You have to have a hot chick as your main character!" When Grand Theft Auto was succesful, it was "You have to let the player be a bad guy!" or "You have to have a free-roaming environment!" (The industry still hasn't gotten over the "Why is GTA so succesful?" phase yet, and we probably won't until GTA gets unseated by something bigger.) The problem is, when you're dealing with these insanely popular games, these games that redefine our ideas of what good game sales are, you can't boil it down to just one factor. It's usually a combination of factors, and together that combination is crazy. Combo!