Top 20 Algorithm
Printed From: Progarchives.com
Category: Topics not related to music
Forum Name: I Have A Question For You......?
Forum Description: Ask any question on any subject: if the admin team or any of our members can answer it we will.
URL: http://www.progarchives.com/forum/forum_posts.asp?TID=56042
Printed Date: May 18 2024 at 10:09
Software Version: Web Wiz Forums 11.01 - http://www.webwizforums.com
Topic: Top 20 Algorithm
Posted By: Epignosis
Subject: Top 20 Algorithm
Date Posted: March 01 2009 at 16:33
|
Hi "PA's I Have a Question for You"-, long time listener, first time caller. I was simply curious about how the top 20 albums are calculated. I understand how ratings are computed, but not how the top 20 albums of each sub genre are determined. I notice that (as it probably should be), an album can have a few hundredths of a point less than another album and still be higher depending on the number of reviews. The math geek within me, who rarely comes out to visit, has. 
------------- https://epignosis.bandcamp.com/album/a-month-of-sundays" rel="nofollow - https://epignosis.bandcamp.com/album/a-month-of-sundays |
Replies:
Posted By: Dean
Date Posted: March 01 2009 at 18:09
|
I think it's something like this ...
(Bayesian Weighting):
br = ( (avg_num_votes * avg_rating) + (this_num_votes * this_rating) ) / (avg_num_votes + this_num_votes)
Which means that an album with a few ratings can be lower in the chart than an album with more ratings and a lower average score. ------------- What? |
Posted By: Easy Livin
Date Posted: March 02 2009 at 02:39
|
If you click on one of the top album lists (on the right hand side of the home page), there is this brief explanation:
|
Posted By: Epignosis
Date Posted: March 02 2009 at 07:53
|
Thanks gentlemen. This will come in handy on some personal projects of mine. Like taking over Antarctica!  Seriously- I learned something today.  ------------- https://epignosis.bandcamp.com/album/a-month-of-sundays" rel="nofollow - https://epignosis.bandcamp.com/album/a-month-of-sundays |
Posted By: The Quiet One
Date Posted: March 02 2009 at 12:41
^I wonder how does that algorithm, helps to take over Atarctica?  
|
Posted By: Dean
Date Posted: March 02 2009 at 12:46
^it helps if you know the mean penguin density ratio - without that the algorithm is useless as the mean penguins will quickly over-power you and your invasion forces. Tsk! I thought everyone knew that.  ------------- What? |
Posted By: Epignosis
Date Posted: March 02 2009 at 12:48
It is a complex plan involving the sacking of Argentina first.  ------------- https://epignosis.bandcamp.com/album/a-month-of-sundays" rel="nofollow - https://epignosis.bandcamp.com/album/a-month-of-sundays |
cacho wrote:Posted By: The Quiet One
Date Posted: March 02 2009 at 12:49
|
^Let me take a few more reads, till I understand that........ ......'2000 years later'...... Nope, still don't get it |
Posted By: The Quiet One
Date Posted: March 02 2009 at 12:51
 Take it! I'm dutch after all! muahahahaha |
Posted By: Epignosis
Date Posted: March 02 2009 at 12:56
We could team up, then! You must bring chocolate and beer, however.  ------------- https://epignosis.bandcamp.com/album/a-month-of-sundays" rel="nofollow - https://epignosis.bandcamp.com/album/a-month-of-sundays |
Posted By: The Quiet One
Date Posted: March 02 2009 at 12:57
| ^and you bring me a physical copy of Epignosis' Still the Waters, and it's a deal! |
Posted By: Epignosis
Date Posted: March 02 2009 at 13:10
Sadly it doesn't exist yet. I'd need $1000 for 1000 professional grade copies with a colored insert and all. I was seriously going to wait until I could do that before making it widely available, but I decided it would never happen, knowing my saving skills. ------------- https://epignosis.bandcamp.com/album/a-month-of-sundays" rel="nofollow - https://epignosis.bandcamp.com/album/a-month-of-sundays |
Posted By: Mr ProgFreak
Date Posted: March 02 2009 at 13:17
|
As pointed out in another thread a while ago: The documentation should link to this page instead: http://en.wikipedia.org/wiki/Bayesian_average - http://en.wikipedia.org/wiki/Bayesian_average ------------- https://tagyourmusic.org/users/Mike" rel="nofollow - https://tagyourmusic.org/users/Mike |
Posted By: crimson87
Date Posted: March 03 2009 at 17:55
|
I don't know if this has anything to do with the algorythm but I ve realized that some Miles davis albums like Kind of blue an bitches brew were at the top of the list a couple of weeks ago ( Specially Bitches that was in the top 25) and after one harsh review it dropped down up to 166 th position more or less.
Has anyone noted this?
|
Posted By: Desoc
Date Posted: April 10 2009 at 02:20
|
^ Yes, this is because the collaborator's ratings are weighted outrageously high when calculating album score. It has nothing to do with the algorithm, but with how each album's rating is calculated. Once this is established, the algorithm is then used to position the individual albums in relation to each other.
Another example of the collaborator bias is this:
http://www.progarchives.com/album-reviews.asp?id=12217 - http://www.progarchives.com/album-reviews.asp?id=12217
which figured in the top 100 list for some time, then because one collaborator gave it a 1-star rating (without review or explanation, mind you) it fell as a rock.
|
Posted By: Mr ProgFreak
Date Posted: April 10 2009 at 02:50
|
^ I think that since this website favors reviews over ratings - and ratings without reviews aren't even visible on the album page - all ratings without reviews should have the same low weight, regardless of whether they're submitted by a collaborator.
------------- https://tagyourmusic.org/users/Mike" rel="nofollow - https://tagyourmusic.org/users/Mike |
Posted By: Desoc
Date Posted: April 10 2009 at 02:54
| ^ On this I agree. |
Posted By: Dean
Date Posted: April 10 2009 at 04:03
|
^ so do I ------------- What? |