One of the best gaming mags I ever read prided itself on the fact that it never ever included a rating for anything. Their logic was that no scoring system can adequately capture a game's nuances, or the way that different people will see different things in a game.
For what it's worth, I agree. I generally can't stand RTS games -- If you asked me to rate Starcraft's value to me out of 10, I'd give it a 2 or 3. I suspect most RTS fans would disagree.
A single number between 0 and 10 doesn't even begin to allow for the fact that some people like RTS games and some don't.
The trick to writing a good review is to not just go on and on about how good or bad you think a game is (although it never hurts to give a subjective opinion at the end). A good review discusses a game's strengths and weaknesses, as well as identifying features that some people will find good and some will find bad (eg the darkness thing in Doom 3 -- it's atmospheric, but many people find it totally annoying). The goal is to give the reader enough info to decide whether a game seems worth trying
for them.
When it comes down to the rating, I try to rate from the point of view of someone who already likes the genre of game in question. Normally that's easy enough to do -- if I didn't like something, I wouldn't be playing and reviewing it.
But at the end of the day the rating is just a number picked by one person, incorporating their opinion of the relative importance of the game's strengths and weaknesses. All it really tells you is whether the reviewer liked it or not -- the reviewer can't tell whether you'll like it. If you want to try and work that out in advance, you need to read the review and make up your own mind.