Review #1 "This game is great because ..." Review #2 "This game sucks because .."
For the users this would make things much easier as they will not need to read 10+ reviews, but just these two (which should increase in quality over time).
In our own sandbox testing, this has worked out well and produced condensed and useful reviews. Obviously though we have an inherent bias - we "want it to work" and thus stay civil and arent starting edit wars. So wondering, if this is something that would actually work in the real world? Considering that an idea as wacky as Wikipedia actually turned out to be very succesful.
Maybe we need some ground rules to have any chance. Suggestions?