Here's what I think it is:

Layout - 45%
Eyecandy - 25%
Fun Factor - 20%
Originality - 10%

More or less. However, it is rare you find a level that has poor eyecandy but a great layout. Sometimes you will find a level that has fairly good eyecandy but poor layout, though. Originality is a different matter.

Things like flow are a combination of multiple aspects. Flow consists about 20 out of 45 percents in the layout and about 5 out of the 20 percents in fun factor. Thus, it is about 25 percent overall. There is a certain amount of eyecandy that goes into the fun factor (5%) and originality as well (10%). The remaining 5% of originality is made up by miscellaneous factors such as music, which probably makes up about 2% of my final rating. Not enough to sway anything.

There is more or less a barrier it seems that keeps levels from going over 8.0 without really great originality. 8.5 would be the absolute maximum for just a standard, little-new level.

I couldn't imagine that MidN would be offended by what FS said; it is more or less accurate. Her levels generally do focus more on originality and fun factor than the more technical stuff (layout and eyecandy.) I'd think that MidN's levels would score high in originality and fun factor (30%), fairly well in eyecandy (25%), and middle-of-the-road in layout (45%). This is an odd case where my rating would be slightly higher than the level would merit; I make subtractions for some technical things. MidN's levels are well-done, but not professional-feeling. This is worth some.

This isn't scientific or anything, but is a more or less accurate measure of how I find that I rate. In reality, though, I simply look at a level for about thirty seconds to a minute and come up with a rating. The more one reviews, the easier it is to judge ratings. I generally have a good idea within my mind at this point, within 0.2 or 0.5, with 0.7 or 1 at the maximum.