There's a lot of dissatisfaction regarding how games are reviewed coming from a number of quarters; there is an equally vociferous defence of the typical numerically-based reviews. Over at GameSetWatch, Simon Parkin takes up the issue of the reviewer-reader divide, especially in terms of what readers want out of a review (even if they don't know it):
The average reader (even if they don't know it) is after a complete objective, scientific comparison between game x and game y with data and statistics and, finally, a numerical point on a linear scale by which they can compare, for example, Mass Effect with Rock Band and see which one is empirically better.
Except, of course, video games don't work in the same way as toasters or digital cameras. Sure, they have mathematical elements and measurable mechanics and it's possible to compare the number of polygons between this one and that and spin out ten thousand graphs detailing how two specimens compare. But, unlike with the Canon EOS400D, I would have no idea at the end of those 25 pages which game was better or where they would sit on the 'true' scale of quality.
On the same issue, the Taipei Gamer proposes a new system that will more evenly balance aesthetic issues that aren't directly related to game play and the interactive elements that are unique to games:
... I am proposing the use of a system which consists of two scores, one measuring the game's excellence from a ludological perspective and the other rating the narrative as it applies to the game. For lack of better terminology I will refer to these as the L-Score and N-Score, respectively ....
The L-Score is the score which is most closely related to the uniqueness of the medium .... Mechanics, systems, and level design are the key components measured by the L-Score.
If the L-Score is a measure of a game's design, then the N-Score is a measure of its artistic achievement. The narrative, in this case, is defined rather broadly. It consists of the game's music, writing, visual style, sound design, overall setting, etc.
I think there are plenty of people who prefer the type of reviews that lend themselves to flash decisions (certainly easier to do when you have a number versus a bunch of text); I don't think the number necessarily needs to go, just perhaps the importance should be mitigated by a different kind of writing.