Forum Home Forum Home > Site News, Newbies, Help and Improvements > Internal news
  New Posts New Posts RSS Feed - TOP 100 Progressive Music Albums
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

Topic ClosedTOP 100 Progressive Music Albums

 Post Reply Post Reply Page  <1 23456 12>
Author
Message
Desoc View Drop Down
Forum Senior Member
Forum Senior Member
Avatar

Joined: December 12 2006
Location: Oslo, Norway
Status: Offline
Points: 216
Direct Link To This Post Posted: July 24 2007 at 05:36
I completely agree, but it would be an easy "mistake" to use the total number of albums in the database, and not only the number of albums with ratings, as a basis for calculating the average rating of all albums. Hence the question.
Back to Top
eduur View Drop Down
Forum Newbie
Forum Newbie


Joined: December 06 2005
Location: Netherlands
Status: Offline
Points: 11
Direct Link To This Post Posted: August 01 2007 at 02:57

Mike thanks for your explanation on the 23th of kuly.
But I still don't get it!

I see how you compute "the average" but for whom ??

With your example with two albums:

================

N=10
R=3.80

now we have an album which has only one rating of 5.0:

n=1
r=5.00

The weighted average is:

avg = (N*R + n*r) / (N+n)

In this case:

avg = (10*3.80 + 1*5) / (10+1) = 43/11 = 3.91

====================

That's nice, but what average? It's the overall average for all albums.
But who would be ranked higher in this case? album (N,R) or (n,r) ?
And why ? (Formula, please :} )

I hate to tell you I have a PhD in Mathematics but it still doesn't make sense to me.

Back to Top
MikeEnRegalia View Drop Down
Special Collaborator
Special Collaborator
Avatar
Honorary Collaborator

Joined: April 22 2005
Location: Sweden
Status: Offline
Points: 21156
Direct Link To This Post Posted: August 01 2007 at 07:00
^ You got it all wrong .. N/R does not stand for a single album:

N = Average number of ratings for all albums in the database. So if there are 100 albums and a total of 1000 ratings in the database, N would be 10.
R = Average rating of all the albums in the database. This is a traditional arithmetical mean, like the avg(...) aggregate function in SQL does.

So N and R are constant values in this algorithm - they're the same for all albums (during the calculation of the weighted average ... of course when new ratings are submitted or existing ratings edited, these constants change).

Now for any given album we first calculate these constants and then apply

avg = (N*R + n*r) / (N + n)

for

n = the number of ratings for the given albm
r = the average rating of the album (arithmetical mean*)

* this has nothing to do with the calculation of this weighted average ... but for the sake of completeness it should be mentioned that the mean value used here is also a slightly "tweaked" arithmetical mean calculation ... for each rating a weight is applied depending on the rating (no review text / review text / collab).
Back to Top
MikeEnRegalia View Drop Down
Special Collaborator
Special Collaborator
Avatar
Honorary Collaborator

Joined: April 22 2005
Location: Sweden
Status: Offline
Points: 21156
Direct Link To This Post Posted: August 01 2007 at 07:13
Incidentally: Another good way to understand what this algorithm does is this:

Suppose that in the archives each album has 10 ratings on average, and the average rating of all albums is 3.5. The average for any given album is then calculated as if the album had 10 additional ratings of 3.5 stars. So new albums with only one rating of 5 stars get "dragged down" towards 3.5 stars ... new albums with one low rating get dragged up towards 3.5 stars. Albums with 10 ratings end up halfway between their own average rating and the total average rating ... and albums with much more than 10 ratings are not really affected, their own average rating remains unaffected.
Back to Top
eduur View Drop Down
Forum Newbie
Forum Newbie


Joined: December 06 2005
Location: Netherlands
Status: Offline
Points: 11
Direct Link To This Post Posted: August 01 2007 at 11:21

Mike, I now understand thanks!

So, in the above case where there are two albums with one album

n1=10
r1=3.8

and the other

n2=1
r2=5

N = 11/2 =5.5
R = 43/11=3.91

album1 = (N*R+n1*r1) / (N+n1) = (5.5*3.91+10*3.8) / (5.5+10)= 3.83
album2 = (N*R+n2*r2) / (N+n2) = (5.5*3.91+1*5)    / (5.5+1) = 4.08

Is this correct?

And do you know R and N for Progarchives...

Back to Top
digdug View Drop Down
Forum Senior Member
Forum Senior Member
Avatar

Joined: July 13 2005
Location: Canada
Status: Offline
Points: 4707
Direct Link To This Post Posted: August 07 2007 at 08:42

Possible Ratings algorithm error ???

I looked at the top 100 for Experimaental/Post Rock
and did not see
Le Notti Difficili by  Lazona..... This album has a rating of 3.36 based on 11 ratings....
 
The top 100 includes many albums with a rating less than 3.36 and less ratings.
 
Please explain.....
Thanks
Prog On!
Back to Top
Desoc View Drop Down
Forum Senior Member
Forum Senior Member
Avatar

Joined: December 12 2006
Location: Oslo, Norway
Status: Offline
Points: 216
Direct Link To This Post Posted: October 07 2007 at 12:19
Originally posted by digdug digdug wrote:

Possible Ratings algorithm error ???

I looked at the top 100 for Experimaental/Post Rock
and did not see
Le Notti Difficili by  Lazona..... This album has a rating of 3.36 based on 11 ratings....
 
The top 100 includes many albums with a rating less than 3.36 and less ratings.
 
Please explain.....
Thanks
 
Seems like a valid point... More people than digdug would like an answer to this Smile ..out of curiosity and search for knowledge, of course. Perhaps Mike could provide an answer given his record in this thread? Wink
Back to Top
progressive View Drop Down
Forum Senior Member
Forum Senior Member


Joined: October 08 2005
Location: Finland
Status: Offline
Points: 366
Direct Link To This Post Posted: February 24 2008 at 14:02
Now it's much much more difficult to search new band when there's

"Only albums with at least 25 ratings and an average ratings over 3.5/5 are displayed in theses lists"

, so eg Zeuhl has only 11 albums in the list and prog folk 30. Why can't people decide themselves how to set the list?? You got the material, large amount of reviews, why can't it be utilized progressively?
DeadCryAngryConfusedShocked

Thumbs%20Upbut.. good luck

Edited by progressive - February 24 2008 at 14:03

► rateyourmusic.com/~Fastro 2672 ratings ▲ last.fm/user/Fastro 5556 artists ▲ www.progarchives.com/Collaborators.asp?id=4933 266◄
Back to Top
Ricochet View Drop Down
Special Collaborator
Special Collaborator
Avatar
Honorary Collaborator

Joined: February 27 2005
Location: Nauru
Status: Offline
Points: 46301
Direct Link To This Post Posted: February 24 2008 at 15:10
I think we should create multiple topics about the new system, whether we wanna write a positive or negative opinion on it.

That is, we can use this one.
Back to Top
dalt99 View Drop Down
Forum Senior Member
Forum Senior Member
Avatar

Joined: April 23 2005
Location: United States
Status: Offline
Points: 454
Direct Link To This Post Posted: February 24 2008 at 23:19
What was wrong with the way it was? I don't understand the change. Why not add an option of what we want to see? If we want to see "Only albums with at least 25 ratings and an average ratings over 3.5/5 are displayed in theses lists", than fine, but we should also be able to CHOOSE to see ALL albums in a genre/studio album/year, etc. of all rating number. Even albums with no votes could be seen. This list was one of the best ways to find out about new album (and more unknown) albums that had only a few ratings but were very good.
Best of 2006 that I've heard:
PFM-Stati Di Immaginazione
Zenit-Surrender (Best "unknown" album)
Oaksenham - Conquest of Pacific
2007:
Phideaux - Doomsday Afternoon
La Torre Del Alchimista - Neo
Back to Top
Harkonnen View Drop Down
Forum Newbie
Forum Newbie
Avatar

Joined: November 12 2007
Location: New Zealand
Status: Offline
Points: 8
Direct Link To This Post Posted: January 06 2009 at 20:10
Dear all,
 
I was looking at the imdb movie database and came accross the "True Bayesian Estimate" to rank the best 250 movies. I think that such method could work well in our PA database.
 
BTW, I just love the flexibility to choose exactly what you want on the Top 100/250 section.
 
I normally choose the studio and live albums with a score higher than 3.8 and more than 50 reviews. I copy the results to Excel and make a table. I normally add a column where I multiply the score times the number of reviews... Not very elegant.
 
I just did the TBE calculation and I like the results... Please give it a try:
 
WR = (vR + mC) / (v + m)
 
Where:
 
WR: Weighted Rank (Score, 0-5)
v = number of reviews
R = average score for that particular album
m = minimum number of reviews to be considered (In my case, 50)
C = average rating of the whole database
Back to Top
Easy Livin View Drop Down
Special Collaborator
Special Collaborator
Avatar
Honorary Collaborator / Retired Admin

Joined: February 21 2004
Location: Scotland
Status: Offline
Points: 15585
Direct Link To This Post Posted: January 07 2009 at 02:50
I think our algorithm is similar.
Back to Top
615694 View Drop Down
Forum Newbie
Forum Newbie


Joined: January 18 2009
Location: INDIANA
Status: Offline
Points: 3
Direct Link To This Post Posted: January 19 2009 at 20:26
Wondering why "Close to the Edge" isn't listed in the Top 100 or did I miss it?
 
It's one of my "5 albums marooned on an island" picks.
"Helm, full dive on the planes."
Back to Top
Moogtron III View Drop Down
Prog Reviewer
Prog Reviewer
Avatar

Joined: April 26 2005
Location: Belgium
Status: Offline
Points: 10616
Direct Link To This Post Posted: January 20 2009 at 02:23
Originally posted by 615694 615694 wrote:

Wondering why "Close to the Edge" isn't listed in the Top 100 or did I miss it?
 
It's one of my "5 albums marooned on an island" picks.
 
It's at nr. 3 Wink .
Back to Top
maru19206 View Drop Down
Forum Newbie
Forum Newbie


Joined: January 20 2009
Status: Offline
Points: 7
Direct Link To This Post Posted: January 20 2009 at 07:53
Again, how we know N and R, number of all in the database?

Back to Top
Dean View Drop Down
Special Collaborator
Special Collaborator
Avatar
Retired Admin and Amateur Layabout

Joined: May 13 2007
Location: Europe
Status: Offline
Points: 37575
Direct Link To This Post Posted: January 20 2009 at 08:04
^ they are dynamic numbers it is not possible to know them because they keep changing as more ratings are posted. They are calculated internally when the chart positions are calculated and are not displayed. I doubt that even M@X knows what the precise values are at any one time.
What?
Back to Top
maru19206 View Drop Down
Forum Newbie
Forum Newbie


Joined: January 20 2009
Status: Offline
Points: 7
Direct Link To This Post Posted: January 21 2009 at 06:54
Ah, yes I got it. Thanks a lot.

Keep on Proggin'.
Back to Top
M@X View Drop Down
Forum & Site Admin Group
Forum & Site Admin Group
Avatar
Co-founder, Admin & Webmaster

Joined: January 29 2004
Location: Canada
Status: Offline
Points: 4028
Direct Link To This Post Posted: May 15 2009 at 02:10
Many changes during the past days to the TOP PROG ALBUM tool and the sub-genre list of key albums.

Basically , the algorithm is mostly the same, it's the default value use in the calculation that have changed.

Here the updated info describing the calculations variables and filters in the TOP PROG ALBUM tool:

Here is some details about how we calculate the average rating of an album and the rank of an album.

  • Average rating: The classic calculation of the average but more weight is affected to the rating of progarchives.com collaborators and to rating with reviews.
    - Rating only: Weight = 1
    - Review by members : Weight = 5
    - Review by PA Collaborators : Weight = 10
  • Rank of an album: We use a calculation that compare an album average rating and number of ratings over all others albums in the query using this theory (http://en.wikipedia.org/wiki/Weighted_average)

  • By default, only albums with at the number of ratings above the average number of ratings in the query and only albums with at the average ratings above the average ratings in the query are displayed in theses lists



The results have changed in both the PA TOP 100 PROG  albums and in the sub-genre too because of the changes in the default variables.

Please post your sugg. and comments here.
Prog On !
Back to Top
Ivan_Melgar_M View Drop Down
Special Collaborator
Special Collaborator

Honorary Collaborator

Joined: April 27 2004
Location: Peru
Status: Offline
Points: 19535
Direct Link To This Post Posted: May 15 2009 at 10:58
I believe that for a bunch of people who follow an unpopular genre which rarely is presented in charts, we care damn too much for the top list.
 
When i invite somebody to PA I don't tell him/her "Check the ratings", I tell them "Read the reviews".
 
Iván
 
            
Back to Top
M@X View Drop Down
Forum & Site Admin Group
Forum & Site Admin Group
Avatar
Co-founder, Admin & Webmaster

Joined: January 29 2004
Location: Canada
Status: Offline
Points: 4028
Direct Link To This Post Posted: May 15 2009 at 11:14
I understand your point IVAN but in the end, we must "anyway" offer some kind of charts as a prog rock web site. And now, with the update algorithm and default value, I believe it's more accurate.

Opinions ?



Edited by M@X - May 15 2009 at 11:15
Prog On !
Back to Top
 Post Reply Post Reply Page  <1 23456 12>

Forum Jump Forum Permissions View Drop Down



This page was generated in 0.219 seconds.
Donate monthly and keep PA fast-loading and ad-free forever.