Best Ever Albums

Re: Chart Manipulation

albummaster — Sat, 19 Jun 2010 04:16:30 GMT

Author: albummaster

Posted: 06/19/2010 08:16

cartoken wrote:

okay, understood the points system. i don't really agree someone who have a 3picks chart should have the same amount of point for his n°2 as a 50 picks chart member.

2nd ranked album in 50 entry chart - 98 points
... in a 3 entry chart - 66 points

A chart with 50 entries *does* contribute more points for a second placed album. Just for the record....

Re: Chart Manipulation

cartoken — Wed, 16 Jun 2010 13:15:16 GMT

Author: cartoken

Posted: 06/16/2010 17:15

okay, understood the points system. i don't really agree someone who have a 3picks chart should have the same amount of point for his n°2 as a 50 picks chart member.

the difference is that the 50 picks chart member will have an extra of points to allocate (94+92+90+88+86+85...etc), which is i think makes it fair for someone who made the effort to make a 50 picks chart.

about the Bayesian average, it was a really good idea !

Re: Chart Manipulation

cartoken — Wed, 16 Jun 2010 13:04:34 GMT

Author: cartoken

Posted: 06/16/2010 17:04

i

Re: Chart Manipulation

albummaster — Wed, 16 Jun 2010 07:47:35 GMT

Author: albummaster

Posted: 06/16/2010 11:47

We've now modified the average rating algorithm so there will be less of an impact if one person decides to rate albums outside of the statistical norm. I hope we can now put this issue to bed.

Re: Chart Manipulation

chrisdread — Tue, 15 Jun 2010 10:18:41 GMT

Author: chrisdread

Posted: 06/15/2010 14:18

I have already apologised to Albummaster for the upset caused by my ratings and requested that he remove them all. Honestly, I was a new member just seeing what happened if I played with the system a bit.

However, what I would like to say is that I don't quite get what right certain members think they have in saying what I do or do not like, or how highly or lowly I should rate certain bands & albums. In all truth, it is absolutely none of any of your fucking business what I rate and how I rate it.

Surely the strength of this site is in the fact that people can genuinely express their opinions about music. If I happen to think that the Beatles are hugely over-rated and have never made a good album (and I do), or that Yes are shite, Pink Floyd boring, Michael Jackson a waste of time - then it is my inalienable right, I thought, to say so.

I seriously had no intent to "game" the site, nor to boost the ratings of albums that I particularly enjoy. I was just seeing what happened. It's called experimentation and some of the narrow-minded, knee-jerk, reactionary muso-bigots might just like to try it some time!

Re: Chart Manipulation

albummaster — Sun, 30 May 2010 04:21:50 GMT

Author: albummaster

Posted: 05/30/2010 08:21

Ok, we're part-way through making the background changes to implement Bayesian averages. I just thought it useful to post a tentative chart to show what a difference this makes:

Current top-rated albums::
1. The Very Best Of Otis Redding;"12";"94.5833";"88.4484"
2. Portrait Of A Legend 1951-1964;"15";"91";"87.0346"
3. Kind of Blue;"164";"90.6098";"90.0624"
4. The Dark Side of the Moon;"420";"90.0357";"89.8276"
5. Substance;"14";"90";"86.286"
6. Neu!;"15";"90";"86.4346"
7. Saxophone Colossus;"14";"89.6429";"86.0777"
8. Led Zeppelin IV;"345";"89.6232";"89.3827"
9. Abbey Road;"386";"89.456";"89.2446"
10. Blonde on Blonde;"197";"89.4162";"89.0138"

... using the new algorithm:
1. Kind of Blue;"164";"90.6098";"90.0624"
2. The Dark Side of the Moon;"420";"90.0357";"89.8276"
3. Led Zeppelin IV;"345";"89.6232";"89.3827"
4. Abbey Road;"386";"89.456";"89.2446"
5. Revolver;"418";"89.2584";"89.0674"
6. Blonde on Blonde;"197";"89.4162";"89.0138"
7. Wish You Were Here;"257";"89.3191";"89.0107"
8. OK Computer;"496";"89.1431";"88.9839"
9. Highway 61 Revisited;"244";"89.2418";"88.9207"
10. The Rise And Fall Of Ziggy Stardust And The Spiders From Mars;"210";"89.0238";"88.663"

The numbers after each title: number of ratings; mean avg rating (current metric); Bayesian avg rating (proposed metric).

The new ratings, to my mind, look more representative. A single 'extreme' vote will have less of an impact on the ratings charts (which I think is a good thing).

Bayesian average formula:
(NumRatings / (NumRatings + MinRatings)) * MeanAvgRating + (MinRatings / (NumRatings + MinRatings)) * SiteMeanAvgRating

where:
MeanAvgRating = mean average rating an album has currently received.
NumRatings = number of ratings an album has currently received.
MinRatings = minimum number of ratings required for an album to appear in the 'top-rated' chart (currently 10).
SiteMeanAvgRating = the site mean average rating for an album (currently 81).

Re: Chart Manipulation

albummaster — Tue, 18 May 2010 15:12:32 GMT

Author: albummaster

Posted: 05/18/2010 19:12

In a different thread, we've sort of said that rather than go for bigger lists, we'd allow multiple and different types of lists so that members can publish their best 50 of the decade, year, genre etc. There are a few things that we need to do first to allow the BEA platform to handle this. I'd rather not go too far off topic on this thread, but please see the other thread where this is discussed and respond there if you´d like to debate this further.

Re: Chart Manipulation

RFNAPLES — Tue, 18 May 2010 09:33:10 GMT

Author: RFNAPLES

Posted: 05/18/2010 13:33

100 album charts!

Re: Chart Manipulation

ffudnebbuh — Tue, 18 May 2010 09:14:11 GMT

Author: ffudnebbuh

Posted: 05/18/2010 13:14

I know that this would never happen, but I would like a 75 or 100 album chart.. there are too many albums from 60 years of pop(Defined as anything other than jazz or classical) music to narrow it down to 50..My first ten are definitely in a specific order, but after that the albums in my chart are separately by angstroms and I am constantly pulling albums out and putting others in...A bigger chart would better express my favorites, just thought I would throw that in.

Re: Chart Manipulation

albummaster — Mon, 17 May 2010 19:42:46 GMT

Author: albummaster

Posted: 05/17/2010 23:42

Freddie55 wrote:

I don't think that their single opinion is less important than any one other one person. Maybe if they rank 10 albums instead of 50 they should get 20% of the points divided among the 10 instead of 100% of the points. Why should someone who ranks only 3 albums have far more impact than you do when you rank 50 albums?

I don't think a 3 album chart is anywhere near as influential as a 50 album chart in terms of its overall contribution.

If you have 50 albums in your chart, you are effectively allocating 2,550 rank points. If you have 3 albums, you are only allocating 200 points.

The only thing that is equal is the top-ranked album has 100 points in both charts. The second ranked album in a 50 album chart will contribute more points than the second album in a three album chart (98 vs 66) (and so will the 3rd ranked album - 96 vs. 33.)

Granted, the lower ranked albums in a 50 album chart won't contribute as many points as the higher ranked albums (but that's because they are ranked lower).

Re: Chart Manipulation

ffudnebbuh — Mon, 17 May 2010 16:17:09 GMT

Author: ffudnebbuh

Posted: 05/17/2010 20:17

albummaster wrote:

We need to remove the possibility of the system being gamed by anybody else. Data integrity is central to the whole point of the site.

If you look at some other music sites, it is fairly common to just have a ten point sliding scale (usually increments of half a star). BEA are the exception at the moment as we currently allow quarter stars and this is a point of difference with some other sites. Let me say here though, we'd never go down to a straight five point system (for me, that is purely the domain of shopping sites).

A ten point scale might strike the right balance of allowing most people to accurately grade an album without comprimising too much accuracy. Together with a Bayesian average, data should be much more protected and every vote still counts. We are not discarding anyone's vote (unlike some systems where the scores can be topped and tailed to get to an average).

There are just too many steps from best to worst at the moment and it is now proven that one person can unintentionally cause mayhem (over 500 ratings by one person at less than 10/100). If we don't put anything in place to stop this, it is inevitably going to happen again.

I think that this is a good idea for a system.

Re: Chart Manipulation

Freddie55 — Mon, 17 May 2010 14:36:59 GMT

Author: Freddie55

Posted: 05/17/2010 18:36

cartoken wrote:

20 different ratings is just fine !!

don't agree about your proposition Freddie, some people (the majority) only have in mind their 10 or less favorite albums, which doesn't mean that their N°1 is less important than a N°1 of some one who has a 50 albums chart!

Re: Chart Manipulation

albummaster — Mon, 17 May 2010 14:30:57 GMT

Author: albummaster

Posted: 05/17/2010 18:30

Whilst I understand and respect the point of view, I don't fully buy the argument that a chart with 50 entries should be more authoritative than a smaller chart. Not every big chart is compiled with the same skill, knowledge or intent. We have some people on this site who have a vast amount of knowledge and experience, but we also have some others who might not have the advantage of many years of listening but who want to fill their charts up with music frem the only decade that they have lived through. We have other people who just use their 50 albums to show off some jokes e.g. Rick Astley, David Hasselhof etc...

There is a case, I think, for a chart with a higher member rating having more authority (where the members decide which are the best charts). However, until the current issue with ratings is put to bed, we should not try and change too many things at once.

Overall, I think the current system strikes a fair balance. It is transparent, simple to understand, and whilst not perfect, does a reasonable job of highlighting the best albums.

Re: Chart Manipulation

cartoken — Mon, 17 May 2010 13:03:06 GMT

Author: cartoken

Posted: 05/17/2010 17:03

20 different ratings is just fine !!

don't agree about your proposition Freddie, some people (the majority) only have in mind their 10 or less favorite albums, which doesn't mean that their N°1 is less important than a N°1 of some one who has a 50 albums chart!

Re: Chart Manipulation

Freddie55 — Mon, 17 May 2010 11:46:01 GMT

Author: Freddie55

Posted: 05/17/2010 15:46

I'm in favour of a 10 increment scale.

I also think that ranking scores attributed to albums for appearing in charts should also be reviewed. For example: if someone has a Greatest 10 Albums chart and places an album at 5/10 it receives 50 ranking points. An album which is ranked by 10 members at an average 48/50 on 10 separate charts only receives a total 40 ranking points. I would argue that the opinion of the 10 members who invested the effort to rate 50 albums in their charts should weigh much more heavily than the opinion of one member who only rates 10 albums in their chart. This type of calculation is having a significant impact on the rankings of the Best Albums of 2010. Just something to think about?

Re: Chart Manipulation

ffudnebbuh — Mon, 17 May 2010 11:10:31 GMT

Author: ffudnebbuh

Posted: 05/17/2010 15:10

I think that you are right.

Re: Chart Manipulation

albummaster — Mon, 17 May 2010 10:58:06 GMT

Author: albummaster

Posted: 05/17/2010 14:58

We need to remove the possibility of the system being gamed by anybody else. Data integrity is central to the whole point of the site.

If you look at some other music sites, it is fairly common to just have a ten point sliding scale (usually increments of half a star). BEA are the exception at the moment as we currently allow quarter stars and this is a point of difference with some other sites. Let me say here though, we'd never go down to a straight five point system (for me, that is purely the domain of shopping sites).

A ten point scale might strike the right balance of allowing most people to accurately grade an album without comprimising too much accuracy. Together with a Bayesian average, data should be much more protected and every vote still counts. We are not discarding anyone's vote (unlike some systems where the scores can be topped and tailed to get to an average).

There are just too many steps from best to worst at the moment and it is now proven that one person can unintentionally cause mayhem (over 500 ratings by one person at less than 10/100). If we don't put anything in place to stop this, it is inevitably going to happen again.

Re: Chart Manipulation

Bork — Mon, 17 May 2010 10:49:14 GMT

Author: Bork

Posted: 05/17/2010 14:49

an_outlaw wrote:

If we can vote on this I would definitely go for the option of having 20 ways to rate an album.

10 would still be okay but 5 would make it far to limited. I tend to avoid magazines which have 5 different ratings for stuff

I started out using a 5-level rating system and soon found that it was too limiting. Changed to 10 after that and have found no need for 20, however I do not find adapting a 10-level system to 20 for this site problematic at all. It's just a matter of ignoring all the 5-steps.

Re: Chart Manipulation

Anonymous — Mon, 17 May 2010 10:10:49 GMT

Author: Anonymous

Posted: 05/17/2010 14:10

If we can vote on this I would definitely go for the option of having 20 ways to rate an album.

10 would still be okay but 5 would make it far to limited. I tend to avoid magazines which have 5 different ratings for stuff

Re: Chart Manipulation

ffudnebbuh — Mon, 17 May 2010 09:42:53 GMT

Author: ffudnebbuh

Posted: 05/17/2010 13:42

The only thing that I am worried about is that I don't want people to feel that they can't express their opinions freely(which would be terrible)...if someone truly believes that an album deserves at 5/100, then they should feel that they can do that. As far as I know, this is the only time that this has happened although I am a fairly new member to this site. Maybe we should leave it the way that it is until it starts happening again?