Reply to thread

Message: <blockquote data-quote="Psion" data-source="post: 37670" data-attributes="member: 172">Making my caseIf you had seen what my eyes have seen...Ahem... I really think I need to make my case here. I think that if you don't do something like weight the reviews, the average system suffers.There have been times that people have posted a review and someone has decided that they disagree with the ratings and are sign up just so they can contradict the review they don't like, and they assign it the most extreme score that they can (a 1 or a 5). This compromises the fairness of the system, because the "reactive" score give is not out of an honest assessment of what they think the product is worth, but an attempt to annul the review that they didn't agree with and to bump the average in the direction they want to. I have seen this multiple times.I understand what you are getting at by saying people can write more reviews but not good reviews. However, that is why I suggested using the root of the number of reviews as a baseline. You get little payoff by just writing 2 or 3 extra reviews just to get your point across about one, so as a result it would discourage this sort of reactive behavior... or at least lessen its effects. If more products got more reviews, it might not matter, since it would all average out. But there are really not that many reviews on a lot of products, so I think some artifical compensation is warranted.Okay, on another note... I recommened earlier that you could half the weight of the highest and lowest review. I think Charles' idea of using the root(mean*median) acheives much of the same thing but would be easier to program since you don't have to worry about the sticky issue of how it related to the weighting system I recommended above.I think number of reviews is a simpler metric, and will be sufficient to address the syndrome that I am worried about.</blockquote>

Verification