C26000

Journal

RSS
Share

AEP Calculation

4 May 2006, 03:05

The AEP is a number that is between -20 and 5 that shows how much preference a last.fm user has for his or her top artists. 5 is the max value for the AEP and it takes this value when the user has listened all the 50 top artists an equal number of times. It stands for Anti Exponiential Points, yeah I know very very silly but I can't change it now :)


The AEP Formula:

AEP = 5 - 25 * ( Slope / AverageTop50 )
Slope = (value of the first artist in the chart - value of the 50th artist in the chart) / 50
AverageTop50 = ( The sum of all values in the top 50 ) / 50

Example:
Tracks Listened from each artist in the Top50 chart:
96
83
82
76
72
66
66
64
63
60
59
58
52
51
51
50
50
49
48
47
47
47
46
45
45
44
44
43
41
41
40
40
39
39
38
38
38
37
37
37
37
37
35
35
35
35
35
34
33
33

Slope = (value of the first artist in the chart - value of the 50th artist in the chart) / 50
Slope = (96 - 33) / 50 = 1.26
96 from the first artist
33 from the last artist

AverageTop50 = ( The sum of all values in the top 50 ) / 50
AverageTop50 = (96 + 83 + 82 + 76 + 72 + 66 + 66 + 64 + 63 + 60 + 59 + 58 + 52 + 51 + 51 + 50 + 50 + 49 + 48 + 47 + 47 + 47 + 46 + 45 + 45 + 44 + 44 + 43 + 41 + 41 + 40 + 40 + 39 + 39 + 38 + 38 + 38 + 37 + 37 + 37 + 37 + 37 + 35 + 35 + 35 + 35 + 35 + 34 + 33 + 33) / 50 = 48.36

AEP = 5 - ( Slope / AverageTop50 ) * 25
AEP = 5 - ( 1.26 / 48.36 ) * 25
AEP = 4.35 Done!

If you use windows you can calculate your AEP using the
Last.fm Extra Stats.

If you are in another Operative System or you prefer to not to install anything in your windows, you can use the AEP Calculator developed by davethemoonman.

If you want to discuss the formula, make sure you read this thread
Accepted Submissions
We Don't Have Exponential Profiles

Comments

  • C26000

    because the webservices don't have the full list.

    26 Nov 2006, 07:05

  • _Tarkus_

    What exactly does AEP stand for?

    15 Dec 2006, 05:46

  • C26000

    @tarkus it's kinda silly :) it stands for Anti Exponential Points :) @InflatableMouse No I don't think so, the aep is only for a single profile, it doesn't compare 2 profiles like the taste o meter.

    15 Dec 2006, 20:09

  • mozes_kriebel

    I got 3,58. I like the idea of this group;) I like to listen to a lot of different artists too, although sentenced is a little on top at the moment;) Good enough to join;)?

    21 Dec 2006, 23:06

  • StarBornDK

    It *is* actually possible to see the entire playlist ... link is the following: http://www.last.fm/user/XXXXX/charts/?charttype=overall&subtype=artist Replace XXXXX with username, obviously, so mine would be: http://www.last.fm/user/StarBornDK/charts/?charttype=overall&subtype=artist Found your group, sounded interesting even though I knew I had no hope of joining (1.85), but decided to investigate anyways :) Hope you can use it :)

    16 Jan 2007, 15:08

  • C26000

    last.fm has the full top 500 artist list in the website but not in the webservices (http://www.audioscrobbler.net/data/webservices/), I really need it as a webservice because it would take ages if I extract the info from the website to update the aep for the members of the group, thanks anyway :)

    16 Jan 2007, 17:56

  • maiser

    C26000: Have you tried contacting someone in the staff? They are usually helpful at least :) It takes great initiative making such tools. Your work is appreciated :) I have an idea. Maybe you could just have a image of the number instead of AEP: number ? Example: Instead of= AEP: 4.13 like this= 4.13 Then its up to users what to write everything but the calculated number.

    24 Jan 2007, 15:49

  • C26000

    I'm reading again my last comment, I think that I was talking about my own problems :), what I said isn't really a very serious limitation to implement it in the software. Actually, I have played a bit with the Top 500 and the Last.fm Extra Stats for generating Stats for the Top 500 list. I have already generated my Top500 tag cloud, but I think it's requires many request to the audioscrobbler webservices that I won't probably release it, anyway with the AEP is easier because I only need the Top 500 list (not 500 artist tag info xml files ;) ), probably I will add it for the next version. @maiser About the idea, I also think it's a good idea, but leaving the option to select whether or not you want to to have the 'AEP = ' in the image, Unfortunately I can do anything about it because I didn't developed the web utility, davethemoonman is the autor, maybe you can suggest it to him.

    24 Jan 2007, 18:59

  • maiser

    oh! thats right. thanks for pin-pointing me :)

    24 Jan 2007, 19:39

  • C26000

    where do I say that the aep is a diversity score?

    25 Feb 2007, 06:47

  • taraturg

    in the first sentence on your page: AEP is a measure of how diverse your music tastes are, based on the top 50 artists in your Last.fm profile.

    25 Feb 2007, 20:03

  • C26000

    mmm that is not my page :P

    25 Feb 2007, 20:24

  • taraturg

    in the first sentence on your page: AEP is a measure of how diverse your music tastes are, based on the top 50 artists in your Last.fm profile.

    25 Feb 2007, 20:45

  • taraturg

    no kidding? well then that would explain that

    25 Feb 2007, 20:46

  • Toolplusplus

    Considered doing this based on tags? So you can get a full picture of how eclectic someone's taste is? :]

    2 Mar 2007, 14:33

  • Toolplusplus

    that said you'd have to objectively define subjective things (differences between genres) and thus = extremely complicated code :(

    2 Mar 2007, 14:34

  • C26000

    I consideredi it, but as you say is not something very reliable read this journal

    2 Mar 2007, 15:03

  • nwo

    Hey CS26000, did you consider calculating a real exponential approximation? What you currently do is only a linear approximation based on the slope of the 1st and 50th track and then scaling it to be independent of the number of tracks listened. But only the 1st and 50th track really matter. So what you want would be a function f(x) = a^x + b, which is the best approximation for say the top50 songs. Depending on the base 'a' you can then classify the user's exponential taste. If you are interested I can supply you a solution. Maybe even some example code, if you tell me the programming language you prefer.

    2 Apr 2007, 08:51

  • C26000

    I consider it in first place, but I had problems with the fit, because not all the profiles adjust very well to an exponential function. take a look at this thread if you are interested -> http://www.last.fm/group/We+Don%27t+Have+Exponential+Profiles/forum/32066/_/106611 [quote]But only the 1st and 50th track really matter[/quote] actually all the numbers matter because the AverageTop50 include all of them, take a look at this example --> http://www.sendspace.com/file/9bsam1

    2 Apr 2007, 16:15

  • nwo

    I'll switch my discussion to that thread, as it is more appropriate.

    2 Apr 2007, 22:02

  • Unwicked

    damn..

    4 Apr 2007, 13:20

  • HeyLuke

    Dividing by 50 in the 'slope' part, [i]and[/i] in the average part seems useless to me. So that leaves a sort of weighted slope times -25, plus 5, right?

    25 Jun 2007, 19:46

  • C26000

    yeah you are right, I think that it's useful for making easier to understand the logic of the formula

    25 Jun 2007, 21:06

  • SixShooterSam

    Doing tale measured like LELB does is more accurate for actual grading of listening to a variety. o hwell.

    2 Aug 2007, 03:02

  • C26000

    how do you define accurate in this situation?, do you have any example where the AEP fails to its purpose?, what is LELB?

    2 Aug 2007, 06:02
See all 80 comments
Leave a comment. Log in to Last.fm or sign up (it’s free).