I have the dubious luxury of some spare time this evening, so I decided to finally post this question that's been at me for a while now.

Specifically, I was inclined to agree with the banishment of album ratings which did not have accompanying reviews.  I agreed with others that, if someone has nothing specific to say in defense of a rating, then I don't particularly care what their rating is since it has no context.

But these ratings aren't completely gone, they're just being hidden.  They still factor into the overall rating of an album throughout the archives.  This seems a bit misleading to me.  I also question the veracity of the math in general (although being a liberal arts major, I could probably be easily convinced I'm wrong on this one).  Here's what I mean:

KANSAS "Kansas" ratings (25 entries)

The band listing page shows there were 26 reviews done on the album.  The individual album page shows 25.  13 of these reviews (those with 200 characters or more) are actually shown.  12 of these rated the album with either 4 or 5 stars.

The chart shows 68% of reviewers gave 4 or 5 star ratings (20% 5 stars and 48% 4 stars).

12 / 26 = 46%

12 / 25 = 48%

12 / 13 = 92%

Rating of the 13 reviews listed = 4.08.  This means the remaining 13 reviews not listed (presumably because they had fewer than 200 character reviews) would have had to average a 3.30 rating if the 3.69 listed under the album is correct.  This is a huge difference from the 4.08 average of the reviewers who took the time to explain the reason(s) for their rating.  This also means over half of the ‘no comment’ reviewers had to have given a rating of 3 stars or less, while 92% of the ‘comment’ reviewers gave 4 stars or more. 

As near as I can tell, this either means those who simply picked a number of stars without writing a review were excessively critical of the album (and therefore further underscores why those types of reviews were eliminated), or the formula for calculating the overall album rating is faulty.

DREAM THEATER "Live at Budokan" ratings (37 entries)

Band listing shows 38 reviews, the album listing shows 37.  Only 9 people bothered to write a review long enough to be shown.  Of these, 8 gave the album 4 stars, and one gave 5 stars.  The chart shows 89% of reviewers gave the album 4 or 5 stars.  That means that the ‘no comment’ reviews are obviously still factored into the ratings, since 89% of 37/38 is actually more than the number of commented reviews shown, and all 9 of those show either 4 or 5 stars.  The rating for the album among the reviews shown would be 4.56; instead, it’s shown with a 4.39.  Not as an egregious an error as the Kansas example, but apparently wrong nonetheless.

So I guess the point is that the reviews without comments should either be shown, or shouldn't be counted into the overall rating of an album.  To count them but not show them is misleading.

Also, whatever method is used to show the number of reviews under the band page and differently under the album page is obviously defective and should be fixed.

This is the algorithm used

ORDER BY power(avg_rating, 3)*log((n_ratings+3)*2) DESC

 

Seriously?  Sorry, I'm really not trying to be a jerk or anything, I'm just trying to understand how a rating is arrived at.  I would think most people assume a simple algorithm like -

n_rating = SELECT Avg(rating) FROM whatever_table WHERE DATALENGTH(reviewlength) > 200 or something similar.  I guess I don't follow why there is a need to raise some average to the third power and multiply that by a logarithm value.  Either way, doesn't explain why the number of reviews on the band page and the number on the album page often doesn't match.  Also doesn't answer the question why 'no comment' reviews are not shown but still count in the rating average.

Again, I'm not trying to be difficult, I just feel the rating being advertised should be clearly explained since it doesn't appear to represent what most reasonable people would assume it to represent, namely a simple average of the total number of reviews shown for that album.

ClemofNazareth wrote: I would think most people assume a simple algorithm like n_rating = SELECT Avg(rating) FROM whatever_table WHERE DATALENGTH(reviewlength) > 200 or something similar.

If you only consider the average rating, the top 100 list would be dominated by album which have been rated 5 stars by one single person.

So you have to find a way to combine the average rating with the number of ratings. The algorithm shown above does just that ... I'm using a similar algorithm on my website and I will publish a detailed explanation of how it works, pros & cons etc. by the weekend.

But I agree that this is off topic - your initial question was why ratings without reviews are counted, but not displayed. Well, good question - currently there is no answer, and the situation is a little bit of a mess.

cobb wrote: chopper wrote: This is the algorithm used ORDER BY power(avg_rating, 3)*log((n_ratings+3)*2) DESC   Perhaps you can explain- chopper. I know this is the end of an SQL query and whatever is returned by the query is listed by the algorithm result descending. I can see part of the algorithm is simply an inner query that returns the average rating and another that returns the total number of ratings, the rest is just straight maths on these numbers, but I can't see why everybody thinks this works so well.

MikeEnRegalia wrote the algorithm - I just copied it here so I can't really explain it. I'll leave that to Mike.

MikeEnRegalia wrote: But I agree that this is off topic - your initial question was why ratings without reviews are counted, but not displayed. Well, good question - currently there is no answer, and the situation is a little bit of a mess.

My humble opinnion, let them be deleted!

chopper wrote: cobb wrote: chopper wrote: This is the algorithm used ORDER BY power(avg_rating, 3)*log((n_ratings+3)*2) DESC   Perhaps you can explain- chopper. I know this is the end of an SQL query and whatever is returned by the query is listed by the algorithm result descending. I can see part of the algorithm is simply an inner query that returns the average rating and another that returns the total number of ratings, the rest is just straight maths on these numbers, but I can't see why everybody thinks this works so well. MikeEnRegalia wrote the algorithm - I just copied it here so I can't really explain it. I'll leave that to Mike.

I already mentioned that I am using this also on my own website. I will write a short explanation about how it works, and I'll post a link here when it's ready.

BTW: Maybe I'll offer a list of variations on my website for you to choose ... then some people can choose to use the average only. In any case I will add a page that lists the top 50 albums using 3 or 4 different implementations - then you can all compare the results and decide for yourself which is the best one.

BTW 2: That was also how mailto:M@x - M@x originally chose the algorithm - by comparing the results.

Clem has a good point.

I've noticed the same thing.  Seems to me you get a lot of people who simply put 1 star on something for whatever reason.

Example: 3% of the people who reviewed SEBTP "rated" it as "poor-Only for completionists." Now, I don't know about you, but if there is a Prog fan out there who really thinks that that deserves 1 star, I want to see them explain why. I'd rather see them write "Oh, I just think Close to the Edge is the best Prog album ever, so I gave this a one to make sure it stayed on top of the ratings. " then write nothing.

I think a lot of people are simply giving the Beatle albums one star just to insure they stay out of the top albums (which isn't to say I think a Beatle album belongs in the top 100 prog albums-just an observation).