"Good Review" has become somewhat of an oxymoron these days, well then again they weren't ever really good in the first place. The big problem here is that the majority of reviews are done by a single person, with one narrow minded point of view. If you think about it, how many times have you come across a game that you thoroughly enjoy, but your friends consider to be crap? The point being that unless you tend to have very similar tastes as the reviewer of a game, that review becomes useless to you.
However, if you let everyone score a game, you have those passionate outliers that try to skew the score to fit their image of a games "deserved score". Human nature dictates that this situation is impossible to remedy as everyone likes to believe they are right. Anyway, back to reviews, which I tend to use as an extended preview of a game. I tend to favor game reviews that list the pros of cons of the games first and present the final verdict upfront. By doing so it allows me to glimpse to see if indeed I'm interested in the game and if I am then I can read "in-depth" part of the review.
I like the 5 star system versus the arbitrary numerical approach towards rating games. About the Moelman's suggestion about doing comparative game reviews, I'm not too fond of that. The thing that happens sometimes with those type of reviews is that they just feed this whole idiotic "console war" attitude. I do like when certain features or elements of a game are compared to an already existing game as it does help the reader visualize certain aspects of a game.
Something that I've always thought of is having a game reviewed by a panel of gamers with different tastes. The problem with this of course and the reason major reviewers don't usually do this type of thing is that it means a lot of people need to purchase/play the game and submit their review which takes time. The main idea though is that you'll have a lot of varying opinions from which to draw your own conclusion as a reader.