This sort of thing annoy me

The Framus 5 album has 45% 2 star ratings and Tull have 50% 3 star. Nope, I don't get it but I'm not going to lose any sleep over it.

The formula is explained here:

 http://www.progarchives.com/forum/forum_posts.asp?TID=121086

The implication is that all albums are scored relative to the query average. If an album has many ratings, those ratings will strongly determine its QWR. If an album has few ratings, the QWR will be close to the query average. This is based on the somewhat strange but not totally bonkers idea that everybody who didn't rate the album (but would've rated an album with average number of ratings) would have rated it at query average. In statistics this is a "quick and dirty" method for missing values imputation; many people know that this can be bad and is better avoided, but sometimes they don't take the time and make the effort to come up with something better.

In particular, for albums above the query average, it is good to have many ratings because only in this way they can rise clearly above the query average. However, for albums below the query average, it is better to have fewer ratings, because this means that their QWR will basically be equal to the query average, whereas a large number of ratings cements their lower average status.

In the posted list by Saperlipopette! the query average seems to be about 3.36. The Framus 5 album has QWR very close to this because it has only 11 ratings, which can't drag it down much. The Jethro Tull album has the largest number of ratings, which gives its actual average rating a very high weight, so its QWR is close to it's actual average of 3.31.   

In my view this idea is somewhat reasonable above the average but makes very little sense below it. Chances are the people who came up with it thought that anybody will mainly be interested in what goes on in the higher ranks, i.e., above the average, anyway.

I should also say that I'm not quite sure how the query average is determined (which is said to depend on the actual query), as there is a default (next to last filter) to only include albums with rating above query average, which would actually require to compute the query average before applying this filter (and maybe others!?), but naively one should think that the query is defined by the filters. (Running some small experiments it seems that the last four numerical filters are ignored, and everything that fulfills the other filters defines the query average. This is probably reasonable.)

Edit: Long, totally off topic digression (which I found interesting)

...Columbia Records created confusion surrounding Shock Treatment by releasing multiple versions of the recording. Ellis attempted to clarify the situation by sending a letter to the “Chords and Discords” forum of Down Beat magazine immediately following the magazine’s review of the album:

“Regarding the review of my record Shock Treatment by Harvey Pekar (DB, Sept. 19), I would like to set the record straight on some little known facts in connection with this album. The copy that was reviewed was one about which I am embarrassed and not proud. The story behind this is as follows:

Upon completion of the album, I did the mixing and editing here in California and then sent the finished product to New York. It wasn’t until the album was already released that I heard a pressing. Much to my horror, I found that without consulting me the whole album had been changed around—rejected masters and unapproved takes were used (not the ones which I had selected and edited), the wrong tunes were on the album, unauthorized splices were made which disturbed the musical flow of some of the compositions (beats were even missing from bars), whole sections were cut out, some of these being the high points of the album. Therefore the liner notes, which were done to the original album, do not agree with what is actually on the album, calling attention to solos and high spots which are not there. I’m surprised that this wasn’t mentioned in the review! Also, the wrong personnel is listed on the jacket. When I discovered what had happened I was, naturally, disturbed and asked Columbia to redo the album. They graciously consented and I was able to change the album back to its original form except that I left Mercy Maybe Mercy, which my producer particularly liked, in place of Zim, which I hope will appear in a future album. Unfortunately, they were not able to call back all the thousands of albums which had already been released. However, they did send a note to the reviewers telling them that the copy which they had received was defective, and to please not review it until they received the corrected copy. It looks as if Down Beat didn’t get that letter. In conclusion, let me state that I have no quarrel with Harvey’s review, but I do wish that he or someone else would review the correct album.”..

Is Equipe 84 on PA? 

I won't disagree, but chances are this is hopeless as I suspect nobody active can change this.

In any case, it isn't trivial to come up with a better formula. The way to do it would be to first collect some "formal axioms", i.e., specify what a good formula should fulfill. Based on this one can design one that does it. Can you define what's wrong? Is the only thing you'd like to have fixed that an album with better average and more ratings should always be better than an album with fewer ratings and lower average? This obviously currently doesn't happen below the query average. Fixing just this would probably be relatively simple, even though it may run counter to the basic concept of the current formula. Of course one needs to take care to not introduce other issues if this one is fixed. (I also wonder whether anybody who is interested would argue in favour of the current system telling you that what actually happens is in some sense better than giving more ratings always a positive influence. I wonder whether there is consensus that you are right - maybe not? Obviously the problem isn't me...)