Recommendation rating

Chao · Post by **Chao** » Sun Jan 23, 2005 6:17 pm

It kinda performs a similar function to a standard deviation, but gives what is a more useful result in this situation. It still gives details on the spread of results, but it's not a mathematical model as such, it's a simple boolean per user based on a cut off value for the opinion, so no heavy number crunching, working out a new SD every time someone adds an opinion. So yeah you have the right idea.

And as it's a recommendation system rather than a review system it can combine the star and opinion ratings, a 4 or 5 star would be a recommend, and a 7+ opinion would be. It's whole beauty is in that it isn't a mathematical sort like all the others are.

And AD, personally I would find it useful for the opposite reason, I'd like it to find the gems that aren't easy to find because they are loved and hated. A lot of those videos get mediocre averages and can end up being seen relatively little so the suggestion query wouldn't have much to go on to recommend it.

AbsoluteDestiny · Post by **AbsoluteDestiny** » Sun Jan 23, 2005 6:34 pm

I just don't think you quite comprehend the reality of the situation - probably because you haven't been on the receiving end of opinions.

As someone who's received over 750 opinions I can honestly say that the patterns you describe do not exist. There are no videos that are as equally loved as they are hated because people rarely leave low scoring reviews and those that do are lost as noise. To most people, a low score would be 7s instead of 8s, 9s and 10s. People who think a video is just mediocre also give 7s.

In theory this would be a lot easier to do with the star scale as the anonymity means people are much more honest. However, you will find that your polar love-it-or-hate-it videos when discovered with star scale analysis are actually just plain "love-it" videos on the top 10% and are easy to find. That's one of the reasons why the top star scale and the top 10% are notably different on certain videos.

I just don't think you're going to find anything new based on the review samples we have at the org - particularly because the videos with enough data to be useful are invariably ones that suffer from the positive feedback loop anyway.

pen-pen2002 · Post by **pen-pen2002** » Tue Jan 25, 2005 12:00 am

AD hit it on the head. There is just not that much variation among oppinions. Although it would be interesting to see what videos had the highest SDs, it wouldn't really be usefull.