I'd really prefer a rating system skewed in the other direction - with fewer numbers for the bad games, and more for the good ones so that more difference can be seen between them.
Still - I've got an old (1993) PC Zone magazine by my side, so let's see how they described the percentage-based scores in those days...
90+: Highly original idea or an exceptional, near 'definitive' version of an old idea. Flawless in all departments and crammed to overflowing with long-term playability. Alternatively, the software company took the reviewer to Paris for the weekend.
80-89: Extremely well executed, strong in all/all but one department. High degree of originality.
70-79: Strong idea, well executed. Not necessarily very original. However, weaknesses do not seriously impair playability, and it has a reasonably long life.
60-69: Games falling into the 'if you like this style of game you'll like this if not you won't' category. Plus games strong on idea but weak on presentation or vice versa.
50-59: Seen it all before, take it or leave it. Nothing special to recommend, but not abysmal either.
40-49: Dodgy. Still playable, but probably only the once.
30-39: Seriously weak in idea and execution.
20-29: The game is hopeless or the software company forgot the £50 cash bribe.
10-19: Bin jobs.
0-9: It didn't work. It crashed the PC and got jammed in the disk drive.
And even back then I'd have rather read the review, often for whatever Duncan McDonald felt like going on about at the time. The Worlds of Legend court case was always a favourite of mine.