Best Ever Albums

Re: Amount of votes should affects!

albummaster — Sat, 29 Dec 2012 05:05:01 GMT

Author: albummaster

Posted: 12/29/2012 10:05

sorry, I mean the value of m (minimum ratings)

Re: Amount of votes should affects!

Jabapac — Sat, 29 Dec 2012 04:55:38 GMT

Author: Jabapac

Posted: 12/29/2012 09:55

albummaster wrote:

good point. I'll add this to the list for early 2013.

Sorry, but do you mean that option (because it's already exists), or to increase the "m" into 20?

Re: Amount of votes should affects!

albummaster — Sat, 29 Dec 2012 04:43:09 GMT

Author: albummaster

Posted: 12/29/2012 09:43

good point. I'll add this to the list for early 2013.

Re: Amount of votes should affects!

Jabapac — Sat, 29 Dec 2012 04:32:05 GMT

Author: Jabapac

Posted: 12/29/2012 09:32

albummaster wrote:

the only issue with doing this is that charts don't seem to get as many ratings as albums or tracks, and so the list of top-rated charts would be most affected by this change. 764 charts have more than 10 ratings, but only 353 charts have more than 20 ratings (which reduces the length of the list quite a bit). I think generally people are a bit afraid to rate charts because they are so personal.

Mmm.. but still there the option of choosing a minimum number of ratings (1 to 25).

Re: Amount of votes should affects!

albummaster — Sat, 29 Dec 2012 04:21:40 GMT

Author: albummaster

Posted: 12/29/2012 09:21

the only issue with doing this is that charts don't seem to get as many ratings as albums or tracks, and so the list of top-rated charts would be most affected by this change. 764 charts have more than 10 ratings, but only 353 charts have more than 20 ratings (which reduces the length of the list quite a bit). I think generally people are a bit afraid to rate charts because they are so personal.

Re: Amount of votes should affects!

Jabapac — Tue, 25 Dec 2012 06:36:57 GMT

Author: Jabapac

Posted: 12/25/2012 11:36

albummaster wrote:

m (yes, not n - sorry) needs to be high enough to make a bit of a difference, but low enough that enough items are included in the top-rated charts. I think it also needs to stay as a round number (base 10) to make it tidier, so I'd be inclined to round to the nearest base 10 figure (which is 20).

I think as well, rather than being dependent on just one item, it should be based on the average number of ratings that the top ten rated items have received to make it more tamper proof (& more stable). If 'm' becomes 1% of the ratings that the top ten rated items have obtained, it would become 1% of 1,214 (which is 12, or 10 when rounded). To make this idea work (I'm not against it), we could base this on 2% to make 24 (or 20 when rounded). All items would then need twenty ratings to qualify for a top-rated chart.

Good idea, but might has sometimes an album with a very low number of ratings on the top 10. it's you decide after all!

Re: Amount of votes should affects!

albummaster — Tue, 25 Dec 2012 03:26:33 GMT

Author: albummaster

Posted: 12/25/2012 08:26

Jabapac wrote:

Yeah, I was thinking of this as an alternative solution, even it's not as much as what I'm hoping for, but it's really fine.
So please let me suggest it that way:
The biggest number of ratings for an item /100.
For now, that item is "OK Computer" with 1524 ratings, lets say 1500. 1500/100=15. so the "m"=15, and so on. what do you think? .. It's just a way to determine the value of "m".

Re: Amount of votes should affects!

Jabapac — Mon, 24 Dec 2012 09:07:31 GMT

Author: Jabapac

Posted: 12/24/2012 14:07

What do you think of my last reply?

Re: Amount of votes should affects!

Jabapac — Fri, 21 Dec 2012 06:24:47 GMT

Author: Jabapac

Posted: 12/21/2012 11:24

albummaster wrote:

Maybe the answer isn't to change AV, but instead to increase the value of n.

Re: Amount of votes should affects!

albummaster — Fri, 21 Dec 2012 04:14:26 GMT

Author: albummaster

Posted: 12/21/2012 09:14

I understand what you are saying, but the way things are now is smoother, and I don't think the number of ratings should affect the average in a significant way once they go over a certain point (100 votes at 90 or 1,000 votes at 90 *should* end up at roughly the same average as enough votes have been cast to get a consensus). I agree that ten ratings isn't very much, but it's enough to get a broad idea. Maybe the answer isn't to change AV, but instead to increase the value of n.

Re: Amount of votes should affects!

Jabapac — Thu, 20 Dec 2012 13:03:34 GMT

Author: Jabapac

Posted: 12/20/2012 18:03

Quote:

A high rated album (mean 90%) with 100 ratings:
(100 ÷ (100 + 10)) × 90 + (10 ÷ (100 + 10)) × 76 = 89%

But the 100 ratings aren't that high on BEA:
(100 ÷ (100 + 10)) × 90 + (10 ÷ (100 + 10)) × 76 = 89%
(500 ÷ (500 + 10)) × 90 + (10 ÷ (500 + 10)) × 76 = 90%
(1000 ÷ (1000 + 10)) × 90 + (10 ÷ (1000 + 10)) × 76 = 90%

So, for the 100 ratings Vs. (whatever a big number was) of ratings.. there will always be only "1" number in difference: 89% to 90%

And don't forget it was with the average of 90, which is far away from the AV: 76. so with closer averages to the AV even the less the differences that will gonna be produced.

But with the AV of 50:

Quote:

(100 ÷ (100 + 10)) × 90 + (10 ÷ (100 + 10)) × 50 = 86%

(100 ÷ (100 + 10)) × 90 + (10 ÷ (100 + 10)) × 50 = 86%
(500 ÷ (500 + 10)) × 90 + (10 ÷ (500 + 10)) × 50 = 89%
(1000 ÷ (1000 + 10)) × 90 + (10 ÷ (1000 + 10)) × 50 = 90%

Now we got "4" numbers in difference: from 86% to 90%

Quote:

Plugging a lower value of AV lowers the average rating of items that haven't been rated very much

5 ratings aren't enough to represent any item's value at all, I believe. not even 10 ratings.

Quote:

but it's also slightly misleading because AV is not derived from the data on the site

If could be able to be changed for example to MR: Middle Range, which equals 50 in our rating system.

Quote:

& the average ends up much lower than the mean.

50 is fair for both the highly rated & the lowly rated.

Quote:

People with charts might not like their average to be much lower than it really is

In case they got few ratings it doesn't proof anything already.. whatever the AV was 80 or 50 or any number.

Re: Amount of votes should affects!

albummaster — Thu, 20 Dec 2012 11:50:17 GMT

Author: albummaster

Posted: 12/20/2012 16:50

Jabapac wrote:

Don't know how actually, but for now I'm thinking of using the site mean average as an invariable value of 50/100. that will help expanding the formula's impact to cover a way larger numbers of votes.

It's not for the sake of popularity, but just to make our rating system more accurate for the reasons I have mentioned before, like reducing the "dislike" effect that popular items are facing, not necessarily abusing, but because the more popular the more known by "not" fan or interested groups.

Code:

Average Rating = (n ÷ (n + m)) × av + (m ÷ (n + m)) × AV

where:
av = mean average rating an album has currently received.
n = number of ratings an album has currently received.
m = minimum number of ratings required for an album to appear in the 'top-rated' chart (currently 10).
AV = the site mean average rating for an album (currently 76/100).

the site mean average album rating is pretty much stable at about 76%, but let's see what happens if the value is artificially reduced to 50%:

A high rated album (mean 90%) with just five ratings:
(5 ÷ (5 + 10)) × 90 + (10 ÷ (5 + 10)) × 76 = 80%
(5 ÷ (5 + 10)) × 90 + (10 ÷ (5 + 10)) × 50 = 63%

A high rated album (mean 90%) with 100 ratings:
(100 ÷ (100 + 10)) × 90 + (10 ÷ (100 + 10)) × 76 = 89%
(100 ÷ (100 + 10)) × 90 + (10 ÷ (100 + 10)) × 50 = 86%

Plugging a lower value of AV lowers the average rating of items that haven't been rated very much, but it's also slightly misleading because AV is not derived from the data on the site & the average ends up much lower than the mean. People with charts might not like their average to be much lower than it really is, so it might not be very popular (there's a very big difference between the two items when an AV of 50 is used). I think using the site mean, it looks a 'truer' average, and the item with more votes still wins out on top (everything else being equal)

Re: Amount of votes should affects!

Jabapac — Mon, 17 Dec 2012 15:02:16 GMT

Author: Jabapac

Posted: 12/17/2012 20:02

albummaster wrote:

Interesting (but I didn't understand everything). How would you represent the updated formula in mathematical terms? (I'd be interested in seeing).

Re: Amount of votes should affects!

albummaster — Mon, 17 Dec 2012 04:40:36 GMT

Author: albummaster

Posted: 12/17/2012 09:40

Interesting (but I didn't understand everything). How would you represent the updated formula in mathematical terms? (I'd be interested in seeing).

FWIW, the number of votes does affect the average at the moment and is one of the variables in the formula. The reason the average rating exists here is because it is a metric that is not based on popularity (as opposed to the aggregated charts), so it provides an alternate window on things. Increasing the popularity element in the average rating calculation would probably end up reducing the variety of albums in the two charts (because the overall chart and top rated charts will become more similar) (the items with the most ratings are likely to be the items people have added into their charts).

Re: Amount of votes should affects!

Jabapac — Sun, 16 Dec 2012 23:35:26 GMT

Author: Jabapac

Posted: 12/17/2012 04:35

Hayden wrote:

I think it's good the way it is now :)

With our recent formula we will forever facing the problem of increasing the "m" (recently 10) for the items which get averages less than the mean site average, which will always pulls them up and never let them down to the places they deserve + it does them an opposite function: the less votes the higher the rating. but with my suggested formula that won't happen and yet will helping reduce the effect of the abused ratings against the popular items that the recent formula does "somehow" but even better I guess!

Imagine 10 users have agreed to take down the average of some album, and they all gave it 5/100, that will does their purpose, but at the same time the number of votes have been increased 10 votes, now that will be with the album' side and reduce the previous impact. and it's much better than the 3 standard deviations that used in apart right now which doesn't work in all situations.

Re: Amount of votes should affects!

Jabapac — Sun, 16 Dec 2012 23:18:13 GMT

Author: Jabapac

Posted: 12/17/2012 04:18

Defago wrote:

Um... If one album got 200 80's and another one 900 80's, the one which got 900 would still be higher. The mean site average loses importance the more ratings an object gets.

Not really, with such high numbers there wouldn't be such a deference since the "m" of the formula is recently equals 10, but even though the flaw of "if the average is near to the site average" that I mentioned still exists.

Defago wrote:

Besides, popular isn't necessarily better.

It's about the popular facing more dislikers, because the popular gets known by many people, those who are interested of or not, but the "less popular" ones are known mostly by those who are interested only, which makes them get higher ratings.

Re: Amount of votes should affects!

Hayden — Sun, 16 Dec 2012 21:58:53 GMT

Author: Hayden

Posted: 12/17/2012 02:58

Nah..

That formula doesn't mix well with BEA's format...

I think it's good the way it is now 😄

I do understand what you were trying to get across though I don't think the 2 other replies fully understood it...

Re: Amount of votes should affects!

Defago — Sun, 16 Dec 2012 21:46:43 GMT

Author: Defago

Posted: 12/17/2012 02:46

Um... If one album got 200 80's and another one 900 80's, the one which got 900 would still be higher. The mean site average loses importance the more ratings an object gets.

Besides, popular isn't necessarily better.

Re: Amount of votes should affects!

Jabapac — Sun, 16 Dec 2012 20:31:09 GMT

Author: Jabapac

Posted: 12/17/2012 01:31

HigherThanTheSun wrote:

The average rating is already affected by the number of votes it has. The formula is to the right of an album's average rating at the bottom of the page.

This means an album with one vote of 100 will not have an average rating of 100, it would be 78.

Right, but that formula loses it's impact with large number of votes (more than 100 for example). so it needs to add one another formula for the large numbers. for example if we have 2 albums with the average of 80, one has 200 votes and the other has 900 votes, they can't be equal. another flaw with our recent formula is that if the average of an album was close to the site mean average it will does no impact at all.

Re: Amount of votes should affects!

HigherThanTheSun — Sun, 16 Dec 2012 11:55:25 GMT

Author: HigherThanTheSun

Posted: 12/16/2012 16:55

The average rating is already affected by the number of votes it has. The formula is to the right of an album's average rating at the bottom of the page.

This means an album with one vote of 100 will not have an average rating of 100, it would be 78.