Other review sites such as Gamespot
and our own Destructoid
have given the game mediocre to bad scores, based partially on intrinsic design flaws such as the small amount of land available to players and the lack of a true single player mode but also to an extent on the server issues, with Destructoid's Joshua Derecher unable to tell if some of the gameplay problems he experienced were inherent to the game's design or due to the broken connectivity.
These are situations that don't lend themselves to review scores. A 4/10 says "bad game" whereas the text of the review itself says "game with problems, some of which might not still be around this time next week". Strip away the rather startling drop in Polygon's score and the reviewer's change in opinion seems a lot more reasonable- from "this game is great" to "think twice before buying" to "I can't recommend this due to technical issues". And yet it's the review score that anyone who visits the site months and even years from now, assuming Polygon doesn't change their score again once the servers stabilize.
Obviously most game don't have this issue, but I think it serves to illustrate the fundamental problem with condensing a reviewer's often nuanced opinion into a simple number or even a five star scale. In even the best of circumstances it's a compromise, one that often only serves to distract attention away from the writer's opinion and facilitate toxic flame wars.
I've heard from many reviewers both inside and outside the gaming world that they don't like using review scores but feel obliged to do so due to audience expectations. In today's rapid fire internet communication age people seem to need some measure of summary to make quick judgements. Kotaku's current binary "should you buy this?" system may be a decent compromise, eliminating a lot of the potential for squabbling and readers demanding that a reviewer explain the difference in quality between a 9.5 and 9.8 score.
LOOK WHO CAME: