If I were making a game review site today, I would probably do something like the following:
5-Star review system with defined rating levels.
UNPLAYABLE Frequent CTD, hardware incompatibilities, showstopping bugs, infected with UbiDRM, etc. Obvious attempt at charging retail prices for a beta product with a promise to "fix it in post".
SHOVELWARE Works, but probably isn't worth playing even by fans of the genre. Poorly designed and/or a blatant ripoff of whatever was popular last year.
MEH A decent game, but nothing noteworthy. This may be due to minor flaws, a lack of innovation, or simply being designed for such a small subset of players that all twelve of them have already heard of it and sent in their pre-orders. You'll probably be better off renting this one or waiting until it goes to the bargain bin.
GOOD Worth paying full price for. Has noticeable gameplay innovations and/or refinements and actually does what it says on the box. Likely to appeal to most fans of the genre, and possibly others as well.
AWESOME Buy this or you won't understand what your friends are talking about for a month. Well polished and advances the genre with significant innovations. A must-buy for genre fans and others alike.
Reviewer Profiles that include lists of each reviewer's preferred genres and favorite games, so you can decide how relevant their tastes are to your own. These should be easily accessible from the review page itself.
Good/Bad/Other Summary. Gamespy does this already and it's a good idea. Shows you the game's strong and weak points at a glance.
Multiple Viewpoints. Each game should be reviewed by multiple people, with at least one being a fan of the game's genre and at least one not being a fan of that genre. If Strategy Steve says Elemental is awesome, you know it's a good game for strategy fans. If Halo Junkie Joey says it's pretty decent, you know it might be worth checking out even if you don't usually play TBS games.
Average user score including comments and a breakdown of how many people voted for each score.
No Publisher Advertising. This creates conflicts of interest, as we've seen in the past. I'm not sure how else you could make money, but maybe donations or site-branded merchandise would be a way to go. Ads for things gamers want that aren't games is another option. Bawls, Jolt soda, and computer gear for example.
Now I'm starting to wish I knew web design...