Collaborators’ TOP 20

Easy Livin,

If you look at my original suggestion for the TOP 20 (see hyperlink below) you'll see that I discussed the algorithm and mentioned the problem of extreme cases.

http://www.progarchives.com/forum/forum_posts.asp?TID=697

One of the reasons I ruled out just using averages was for the reason you point out. (Another reason is that it does not allow differentiation between the hundreds of albums that would have the same average rating.)

If you look at "Train Of Thought" you'll see that it *is* popular:

33 5-star ratings; 9 4-star ratings; 11 3-star ratings; 5 2-star ratings; 8 1-star ratings.

Compare that with "Dark Side Of The Moon":

28 5-star ratings; 2 4-star ratings; 3 3-star ratings; 1 2-star rating.

So, from the above, to me "Train Of Thought" *is* more popular than DSOTM whichever way you look at it. And addressing your comment about the number of people who have rated it, the fact that more people have rated it is, in itself, a possible indication of popularity.

When I checked the "real" TOP 50 list in June, it was a good reflection of popularity, i.e. the 'extreme case' scenario seems not to occur in practice.

Fun, it is!

But, as I think we both agree, the comparison of averages is problematic:

Album 1 has 20 5-star ratings => average = 5-stars

Album 2 has 60 4-star ratings => average = 4-stars

Which of the above two albums is more popular?

I think the number of reviews needs to be a factor. And I think that the more people/reviews are included, the better (i.e. the bigger the sample, the more interesting and representative the result).

I need to think some more about this (interesting) problem!

MAX@ wrote: "Should we count all ratings by collaborators and non-collaborators ?" MAX@

I still say no to that one Max. I know at first it seems a bit elitist, but I retain my concerns about spamming when it becomes known that all votes count. (And rest assured it would, the bulletin boards on the various band sites would soon have notices on them saying "vote for our band"). Until we collecively come up with a reasonably foolproof way of doing it, I say stick with what we have.

Thanks, MAX@.

Well, regarding your second question, I think my opinion is known, but I also understand Easy Livin's concerns.

Is there some way we could protect against the spamming that Easy Livin thinks may be a possibility? For example, only include star-ratings from non-Collaborators if there is an associated review of at least a certain length? Only disclose the results to members of the forums?

Whatever is done, I think it would still be useful to have the Collaborators' TOP 20/50 as a separate list, if for no other reason than to give some recommendations to newcomers from the 'official' site reviewers. But as we've seen in this thread, the Collaborators tastes do not appear to be an indication of the tastes of the larger population.

When I get time I'm going to think about the algorithm. But I'm convinced that averaging is not the best way to do it: it does not make sense to me if an album has, say, ten 5-star ratings (average = 5 stars) and then someone adds a 4-star rating (average = 4.91) and the album drops in the ratings. Easy Livin might counter by saying that adding e.g. a 2-star review should not raise an album in the ratings, and he might also be right. But if Album1 has more 5-star and 4-star reviews than Album2 (as in the example I quoted in an earlier posting) then, to me, even if Album1 has more 2-star and below ratings, Album1 is the more popular. I say this because more people like Album1 than Album2, simple as that. But I'll give the algorithm some more thought.

JrKASperov wrote: threefates wrote: I don't!!  ELP is way underrated, BSS should be in the top 10 and Trilogy is missing.  That in itself is a sin!! No.

No what?