Put a Pin on the Map View my Forum Guestmap
Free Guestmaps by Bravenet.com

The Old Acclaimed Music Forum

Go to the NEW FORUM

Music, music, music...
Start a New Topic 
1 2
Author
Comment
The list-making workshop : sharing our list-making methods

As I promised you, this is a thread that I hope will be fruitful.

Like good housewives sharing recipees, let us talk shop. We all love music, and we all love to make lists.
Our wives/girlfriends (if ever we have wives and girlfriends or if they haven't (already) ran away ), friends and relatives, when we talk about this strange affection, say "how intersting" but think "weirdos".

That's why we come to a site called Acclaimedmusic.net
That's what the Web is for : gathering niche people like us that couldn't meet otherwise.

Now let's talk about how we make those lists. Some of us (the most sane of the lot?) use the Listyguy method " I just think to myself, "Which album do I like more?" That's probably why I do so much tweaking!"

But I know that some of us use strange calculations. I'd be very interested to share these tricks. I konw it's the case of BIllAdama and Andre (I'd love specifics about your respective ways and means guys).

As I have a lot of time these days (I'm quitting my job in August and I'm sitting here at the office with nothing much that I want to do), I'm gonna explain how I proceed.


Summary (for those who don't want to read it all) : Each song gets a 0-5 rating. The album gets an average rating wih adjustments : the worst song is taken out, bonuses for the good songs, and a comparison bonus/malus because i'm rating albums 3 by 3.



IMO, a good album is first a good collection of songs, even if sometimes the whole is thing is better (ex IMO Blonde On Blonde ))or worse than the sum of its parts


So when I'm listening to an album, I rate the songs. I give them a 0 to 5 rating (in fact I rarely give less than 3, but it can be 3.25, 3.5, 3.75, etc...)

(sorry if this scale is not very clear, but it's a translation from French)

5 = perfect

4.75 = almost perfect (but I don't use it much anymore)

4.5 = excellent (universal)

4.25 = remarkable (excellent, but doesn't go above its zone/genre)

4 = good

3.75 = not bad at all ("pas mal du tout")

3.5 = decent ("pas mal")

3.25 = just fair ("passable")

3 = indifferent

2.75 = not great ("pas terrible")

2.5 = unsufficient

2 = bad

1.5 = very bad

1 = revolting

0 = violates human rights

Generally, a song under 3 is a song I want to stop before the end.

When this is done, remove the worst song (or the 2 worst songs if you've got more than 15 songs, 2 from 16-25, etc... or 0 if you've got less than 6).

Then calculate the average rating

Multiply by 20 (you get a percentage)

Apply bonuses (0.1 for each 4 song, 0.2 for each 4.25, 0.5 for each 4.5, 0.7 for each 4.75, 1 for each 5)

then you've got your rating.

Sometimes to adjust with the overall impression (the whole vs sum of parts thing) I apply bonuses or I "tax" the album. Some songs have small + or - attached to their scores when I'm not sure about a rating, so i apply these - or + to correct the overall score.

So with the bonuses, some albums can get more than 100 points (14 albums in my all-time list have more than 100)

If 2 or more albums get the same score, the following criteria are number of listens then total bonus then standard deviation

Advantages : no tweating anymore, no doubts, list is always ready, no album is forgotten, the rules evolve with time

Flaws : lack of spontaneity (no shit), some albums were rated long ago while the rules were different.
The biggest flaw of the bonus thing is that it favors long albums over short ones. But I love long albums like I love long movies or long books, so that's not a problem.
The rules have been changing since the beginning (when I was infected with the AM virus in 2007), so to reflect my real tastes and preferences.

Another flaw is that there is no comparison (like in AM 3000 or the BIllAdama and Andre methods). But I've introduced something new : Instead of calculating each album's score just after hearing it, I wait until I have heard 3,4 or 5 albums and I make my ajustments by comparing them.

Since the beginning I've rated 1147 albums (of course, some albums got multiple listenings).

That's about it. Still speaking to me ?

I also do some pre-listenings for the decade and EOY polls using more or less the same method but out of 4-5 songs.

Well, well, well, now you can cheer and say whatever the guys say at AA mmetings when someone is done with his speech.

Who's next ?




PS : When making songs list I still more or less use the tweaking method. I'd be interested in a comparison method at least for songs so please don't hesitate to share yours (or to say that the tweaking is better than all this nonsense)

Re: The list-making workshop : sharing our list-making methods

nicolas- I don't have the time right now to detail my entire rating system, but I wanted to mention that it is REMARKABLY similar to yours.

Instead of 0-5, I use 1-10 to score songs, also at 0.25 intervals (ie- 10.0, 9.75, etc.), with "bad" songs falling under 6.0, but almost never under 4.0 (I have yet to rate a Spin Doctors album!)

One difference here is that I don't drop off the low scores, but I do weight for length of song.

I also apply bonuses for 10.0, 9.0+, and 8.0+ songs similar to you.

I have a small calculation at the end that factors in overall length of album, but it's minimal (favors longer albums). Also, I try to account for more nebulous factors, such as album coherence, historical attachment, influence on my life & tastes, and other things within the rating of each song.

I currently have 1264 albums rated, with 3 more lined up to go today.

Great idea for a thread!

Re: The list-making workshop : sharing our list-making methods

I feel less lonely

About song length : If a song is less than 2:00 long, I can chose not to rate it if it doesn't make sense.

About song rating : if I find that a song is good but I feel no "personal involvment" in the listening, it won't get a 4. It's the typical 3.75 !!! A 3.75 can be a good song spoiled by bad lyrics, bad performing or any other flaw.

Re: The list-making workshop : sharing our list-making methods

nicolas
As I have a lot of time these days (I'm quitting my job in August and I'm sitting here at the office with nothing much that I want to do)

Forget about the calculations. THIS is the best method for making great music lists.

Re: The list-making workshop : sharing our list-making methods

do you mean having the time ?
But I love my calculations, even if there are 8 Springsteen albums, 8 Beatles, 9 Brassens, 6 Neil Young, 6 Dylan 5 tom Waits etc in my top 200 (and there it goes again..)

Re: The list-making workshop : sharing our list-making methods

I’m... with... Listyguy on this one, I guess.

After listening to an album in its entirety I give it a score out of 100. It makes complete sense in my mind why an album should get a score of 89 instead of 90 or 88. It probably won’t make sense to anyone else, but here are some examples:

95-100: I love it so much that it should be in my all-time top 50
94: Still love it but it’s juuuuuust not quite top 50
93: Not even thinking about it as a personal top 50 album but still love it enough to give it 5 stars
92: Very, very close to a 5-star album but I can’t quite get myself to calling it a 5-star album
etc...

When albums have the same score, I just follow my gut feeling to rank them. Of course, by listening to an album again I might give it a higher or lower score and it will move up or down my list.

For me the album as a whole is more important than the sum of its parts, which is why I don’t give scores to individual songs when deciding on an overall score. In the 2005 poll that we’re doing at the moment, for example, Arular is my #1 album and Illinois my #2. If I ranked the individual songs of each, Illinois would probably come out on top. The peaks of Illinois are much higher than the peaks of Arular, but I find Illinois a little wearisome at times whereas for me Arular is a non-stop rapturous blend of crunk, garage, dancehall, rap, electronic music and even a bit of Afrobeat (I even like the ridiculous skits). So because Arular engages me throughout it very narrowly beats Illinois to my top spot. The whole beats the sum.

Interesting thread. I look forward to reading more posts. The geek factor on this forum has been a bit low recently, but this is taking it to whole new levels (!).

Re: The list-making workshop : sharing our list-making methods

Well my method is much more like Listyguy's than anyone elses - I just compare two albums, say which one is better than other, then apply numbers to it. So for example, when I listen to The National's Alligator and Boxer, I think to myself "well Alligator is slightly better than Boxer". And so I give Alligator a score of 16 and Boxer a score of 15 (out of 20). Then I think, well High Violet is probably equally good to Boxer so I give it a 15 as well. So then I get a huge list of numbers and I put them in order. The hard job is sorting out between the various 17s and 16s and I just use my heart for that.

Re: The list-making workshop : sharing our list-making methods

nicolas, I used to have a similar method to yours, but changed it recently towards following procedure:

1) Select target list (e.g. "best songs of 1997")
2) Based on your personal music collection and a bit of scanning through AM and RYM rankings, create a nomination list of X (e.g. 200) eligible songs.
3) Create a waiting list, which initially is empty.
4) Randomly pick 5 songs from nomination list
5) Listen to the 5 songs and give scores 1 (worst) to 5 (best).
6) Remark: If I can not listen to a song until the end, it will be removed from list (the song would in the end get rating below 5.5 and listening should be fun, so: out!)
7) Move the selected songs from nomination list to waiting list.
8) As long as nomination list is not empty, go back to step 4.
9) Move all songs back from waiting list to nomination list
10) Sort nomination list from lowest to highest scores
11) Pick 5 songs with lowest score. If multiple songs have same highest score, pick remaining songs randomly.
[Example: When sorted scores list is 1,1,1,3,3,3,3,4,... Then 3 songs with score 1 are picked automatically and from the 4 songs with score 3, another 2 are randomly picked]
12) Listen to the 5 songs and give scores 1 (worst) to 5 (best).
13) Move the selected songs from nomination list to waiting list.
14) As long as nomination list is not empty, go back to step 11.
15) Move all songs back from waiting list to nomination list
16) Sort nomination list from lowest to highest scores
17) If songs have not been listened to Y (e.g. 4) times, go back to step 11.
18) Listen to song with lowest score to set lowest rating: Default is 5.5, but sometimes a rating of 6 is allowed.
19) Listen to song with highest score to set highest rating: Default 9.5, but sometimes a rating of 9 or 10 is allowed.
20) Using the song scores, calculate the song ratings as follows:

rating = lowest rating + score * (highest rating - lowest rating) / (highest score - lowest score)

and round of to half points.

My rating index (Stephan, I recently modified my rating procedure a bit towards yours at RYM)
10.0 = close to perfection
9.5 = masterpiece
9.0 = outstanding
8.5 = excellent
8.0 = great
7.5 = good
7.0 = nice
6.5 = okay
6.0 = mediocre
5.5 = weak

For updating the list I am stealing the concept of the Divisions game:
Each rating is like a division and songs can go up and down a division for each run through.

Hope you were all able to follow this ...

Re: The list-making workshop : sharing our list-making methods

BTW, this procedure is for songs only.

Rating album songs individually to get album ratings would ruin my life (yes, I do have a wife and social life)

So, there I listen to the entire album and stay with the simple listen-and-rate procedure.

Re: The list-making workshop : sharing our list-making methods

Andre


Rating album songs individually to get album ratings would ruin my life (yes, I do have a wife and social life)


I spend about 2h30 in the metro and train everyday on my way to work and back, that's why I have so much time to listen to albums (and that's also why I'm quitting
At home I just listen to albums without rating them. I usually play albums I know I'm gonna rate later. Now i'm playing Antony & the Johnsons cause I think it needs a new spin (it's competing against Sufjan for the 2005 poll). I'll rate it tomorrow in the metro.

Andre I'll try your song method for my all-time song list. I have something like 500 5-star songs but I can reduce it and create a starting list.

Re: The list-making workshop : sharing our list-making methods

I've mentioned my scoring method already, but basically I have a program that works like this.

I'll put in four to eight albums I just listened to, ranked from most to least enjoyable on that listen. The program then takes all the lists I've ever put in, weights them based on both how long ago I made them and how long I had owned the album when I made them, and puts them all together in one list.

Here's in short how this algorithm works:
All albums start with 0 points. Then, an album is 'maximized' by finding the score from 0 to 100 that minimizes its 'Incongruity'. It finds the score with the least 'Incongruity' based on binary search. That is, you start with 50, and check whether 75 or 25 has lower incongruity. If it's 75, then check 62.5, 87.5. Keep repeating this, dividing by half. If the two scores have equal incongruity, it will go to the lower number.

The Incongruity rating is based on this: For every win, add max(other album score - this album score + 10, 0)^1.05. For every loss, add max(this album score - other album score + 10, 0)^1.05.

Then, it repeats this process on all albums in random order, 100,000 times. On each iteration, it reduces the amount an album's score is allowed to change. On the first iteration an album's score can change 100 points. On the last, 0.000001 points.

It's actually a little more complicated than that, but if you can manage to understand this, you will understand enough to make it work for yourself.



Oh, and I have two separate lists. One for my favorite albums of all time, and one only for albums released in the past twelve months. On August 1st I will delete all the albums that I picked up between 8/1/10 and 8/31/10.

Which is these four:
Arcade Fire;The Suburbs;2010;2010;7;4;52.487146
Menomena;Mines;2010;2010;7;4;0.000000
Tame Impala;Innerspeaker;2010;2010;7;25;52.606784
Sufjan Stevens;All Delighted People;2010;2010;7;25;44.304381

Re: The list-making workshop : sharing our list-making methods

For albums, I just open up an excel file and create a file for my all time albums list. I then go through iTunes and listall the albums that may make my top 100 in the file. Then, I listen to each a few times, and make a list. That's my plan, anyway. I haven't gotten too far with it yet. After I have a base list, I figure I can add albums over time.

Re: The list-making workshop : sharing our list-making methods

I, too, have written a computer program to help me rank albums, songs and artists. I work from an Excel spreadsheet which contains fields for artist, album/song, year and star rating (for albums only).

I'll refer to albums for the rest of the post, but songs work the same way, unless pointed out.

The program then chooses 2 albums at random from a given customizable set, and I select which one I prefer. This set can be customized to include all albums, albums from the same artist, albums from the same year, albums with the fewest comparisons, etc. The default is to choose one album that has the fewest comparisons and then match it up against another album within the 1 standard deviation of its points (except early on, when the albums haven't spread out enough in terms of points, in which case the second album is chosen to be within some arbitrary point range, depending on the point system used).

I have been toying around with 3 different point systems:

The first point system I devised uses a weighted average of win percentage, strength of schedule (win percentage of opponents matches against other albums) and initial star rating (this part is left out for songs). It produces results between 0 and 100 and has worked ok, but I found a couple of drawbacks. First, I felt the weights were arbitrary and did not know how to properly tweak them. Second, I found myself forcing certain comparisons in order to "fix" the result.

The second point system I devised is based on the ELO Chess rating method with a k-factor of 24. Each album is initialized to have 1000 points, and then individual matchups add and subtract the same number of points to the winner and loser. "Upsets" (in which an album with a lower number of points beats and album with a higher number of points) make the result of the match weigh more heavily. This method seems to work pretty well, but I find that it takes a looooong time for the algorithm to converge well. It seems very sensitive to the number of comparisons, as albums generally follow in tiers depending on the number of losses they have. With 729 albums in my database and having compared each album to at least 35 others, I still don't think it has come close to converging.

The point system I'm currently working on is based simply on the matrix M of matchup results. I start with an identity matrix and add 1 to the victor's row and loser's column for each matchup. This means that for a particular album, you can look along its row to see the albums it has beaten (and how many times in the case of repeated matchups). The points for each album are the sum of the corresponding row of M*M over the sum of the corresponding row of t(M)*t(M), where t is the transpose. A transformation is then made to center the points at zero and reciprocate the negative scores for symmetry. I like that this method does not rely on arbitrary weights as the above 2 methods do. It's still early going (I've only made 5 comparisons for most albums), but a normal quantile plot of the points for albums with at least 1 win and 1 loss looks pretty beautiful:



I also use kernel density estimation to assign ratings on a 100-point scale to each album based on its rank and the distribution of star ratings.

The result looks like this for albums:

Re: The list-making workshop : sharing our list-making methods

To BillA and Moonbeam,

Your system looks great but it's beyond my math/statistics/computer programming abilities. I did political science as a major !!!

It would be cool to have quch a program. Hopefully Henrik is coming to Paris in a couple of weeks so I'll ask him if it's easy to build such a program.

These programs heve to be custom made I believe because they have to reflect the way and criteria you use to appreciate music.

Re: The list-making workshop : sharing our list-making methods

nicolas
To BillA and Moonbeam,

Your system looks great but it's beyond my math/statistics/computer programming abilities. I did political science as a major !!!

It would be cool to have quch a program. Hopefully Henrik is coming to Paris in a couple of weeks so I'll ask him if it's easy to build such a program.

These programs heve to be custom made I believe because they have to reflect the way and criteria you use to appreciate music.


I'd be happy to share mine, but it gets a bit complicated if you aren't familiar with the statistical software R.

Re: The list-making workshop : sharing our list-making methods

thanks Moonbeam ! I'm definitely interested in that stuff and I have a little time. What would you advise me to get a little more familiar with statistic software ?

It is pretty obvious that relative approach (based on comparisons) is more relevant that absolute approach (like mine, even with corrections)

We'll end up funding a list-making resaerch laboratory !!

Re: The list-making workshop : sharing our list-making methods

nicolas
thanks Moonbeam ! I'm definitely interested in that stuff and I have a little time. What would you advise me to get a little more familiar with statistic software ?

It is pretty obvious that relative approach (based on comparisons) is more relevant that absolute approach (like mine, even with corrections)

We'll end up funding a list-making resaerch laboratory !!


Well, you can download R for free from CRAN. It is a command based software, so learning a few commands wouldn't hurt.

If I end up sharing the program, I'll have to make it a little more "user-friendly" with perhaps a "how to" file and so forth. Let me know if you are interested (and anyone else, too).

Re: The list-making workshop : sharing our list-making methods

Ok thanks Moonbeam, I'll let you know soon.

Re: The list-making workshop : sharing our list-making methods

nicolas
Ok thanks Moonbeam, I'll let you know soon.


Another person has expressed interest in my program. I'm in the process of cleaning it up and writing instructions, and hopefully I can share it soon!

Re: The list-making workshop : sharing our list-making methods

I'd be interested as well!

What I have right now is the most basic system on the planet, a program that simply gives me two albums of which I pick one and these albums then switch places if I tell the program I prefer the lower rated one. It's all I've got in terms of scripting skills, but it's rather prone to flukes. I try to correct this by first entering my top 50, then 25-75, then 50-100, etc., but it's a long process.

Upside is, it's very easy to share (I made it for IRC, so I can even run the bot for you), so if anyone is interested feel free to ask. I'd wait for Moonbeam's though.

Re: The list-making workshop : sharing our list-making methods

Moonbeam

Another person has expressed interest in my program. I'm in the process of cleaning it up and writing instructions, and hopefully I can share it soon!


Then I'll be definitely interested !

Re: The list-making workshop : sharing our list-making methods

Count me in, moonbeam!

Re: The list-making workshop : sharing our list-making methods

...and me :0

Actually, Stephan can I have yours as well?

Re: The list-making workshop : sharing our list-making methods

The current version of mine runs on the Java command line and has a lot of input issues. It's basically the same thing I offered to the forum a couple years ago.

I could try to make a more graphical version of it that's more user friendly, but right now my other priorities may put it off a while.

Although, if I could charge $1 for it, I could call it a commercial product on my resume. Hmm...

Re: The list-making workshop : sharing our list-making methods

OK, I've written a set of Instructions and initialized everything for new users. The question now is how to share it. I don't have a website or blog. Would something like sendspace be the way to go?

Re: The list-making workshop : sharing our list-making methods

Moonbeam, I'm interested too, the sendspace solution would work fine for me.

My math study lies way back in history. I had some experience with the S language, but is this R like a better, evolved extension?

Re: The list-making workshop : sharing our list-making methods

sendspace is fine

Re: The list-making workshop : sharing our list-making methods

OK, here is the sendspace link to a zipped folder with all of the necessary documents.

Let me know if you have any questions. Enjoy!

Re: The list-making workshop : sharing our list-making methods

Moonbeam
OK, here is the sendspace link to a zipped folder with all of the necessary documents.

Let me know if you have any questions. Enjoy!
This is brilliant. Thank you for sharing it with the forum.
Edit: When I try to save the album list, is tells me to make sure the path and file name are correct. I'm not sure why, though.

Re: The list-making workshop : sharing our list-making methods

Thanks Moonbeam !!

I first looked into the notepad and got scared . Then I opened the Word document and everything seems clearer now.

Re: The list-making workshop : sharing our list-making methods

Moonbeam
OK, here is the sendspace link to a zipped folder with all of the necessary documents.

Let me know if you have any questions. Enjoy!


Works like a charm, 1999 was clearly leading!

And the fine-tuning is really super: set = decade(2000) actually only picks albums from the noughties!

I checked the code in the txt file and will for sure explore the details at a later stage.

Thanks a lot, Moonbeam!

Edit - Question: What influence have the stars for a specific album when this album has been in at least one comparison? None? As I understand, the console reads the ratings from the RData file, so it does not really need the stars anymore or does it?

Re: The list-making workshop : sharing our list-making methods

This is exactly what I've been looking for. It's brilliant (although at the moment myfavourite album London Calling is languishing at #139, though I'm sure that's my fault)!

Re: The list-making workshop : sharing our list-making methods

Happy to be of service!

As for the stars, they aren't necessary for the ranking at all. The stars are simply there for the kernel density estimation to provide a Rating for each album from 0-100 to be displayed when you print your list to the screen. The program saves only the details of the comparisons in the RData file (the matchup results, points, wins, losses and the current set) and it reads in the album/year/star details from the text file every time you start the program.

As for your #1 album not being #1 immediately, it should quickly ascend. I find that this point system is pretty excellent for quickly finding the best (and worst) few albums in the list and much better than the previous point system I employed. For example, I have 729 albums in my database, and after each album had been compared at least 10 times, my 1-5 and 725-729 were correctly identified. To check how many comparisons the albums in your list have, you can type tot into the R Console. Type sort(tot) for an ordered list. Albums in the middle regions of the list will take longer to hone in on their natural place than the top and bottom, but I think that's probably true for how I feel, as those albums fluctuate around more in my own perception than the very top and bottom.

Re: The list-making workshop : sharing our list-making methods

Listyguy
Moonbeam
OK, here is the sendspace link to a zipped folder with all of the necessary documents.

Let me know if you have any questions. Enjoy!
This is brilliant. Thank you for sharing it with the forum.
Edit: When I try to save the album list, is tells me to make sure the path and file name are correct. I'm not sure why, though.


Are you editing your album list using Microsoft Excel, or some other software?

I use Excel, and it always prompts me "Album List.txt may contain features that are not compatible with Text (Tab delimited)."

It never seems to be a problem, though. As long as the program is working and all of your albums are there (type albums into the R Console to see your list of albums after exiting the comparison loop).

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Are you editing your album list using Microsoft Excel, or some other software?

I use Excel, and it always prompts me "Album List.txt may contain features that are not compatible with Text (Tab delimited)."

It never seems to be a problem, though. As long as the program is working and all of your albums are there (type albums into the R Console to see your list of albums after exiting the comparison loop).
It ended up saving, I had to "unzip" the file. However, I am now running into a new problem. I cannot view my list.

Re: The list-making workshop : sharing our list-making methods

Many thanks for sharing this, Moonbeam.

*goes back to lurking*

Re: The list-making workshop : sharing our list-making methods

Listyguy
Moonbeam
Are you editing your album list using Microsoft Excel, or some other software?

I use Excel, and it always prompts me "Album List.txt may contain features that are not compatible with Text (Tab delimited)."

It never seems to be a problem, though. As long as the program is working and all of your albums are there (type albums into the R Console to see your list of albums after exiting the comparison loop).
It ended up saving, I had to "unzip" the file. However, I am now running into a new problem. I cannot view my list.


I should have explained that you need to unzip the folder!

In terms of viewing the list, do you mean within R? Are you able to get into the comparison loop? This loop will continue until you press something other than 1 or 2. Once you exit the comparison loop, you can see your ranked list by typing display and hitting Enter.

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Listyguy
Moonbeam
Are you editing your album list using Microsoft Excel, or some other software?

I use Excel, and it always prompts me "Album List.txt may contain features that are not compatible with Text (Tab delimited)."

It never seems to be a problem, though. As long as the program is working and all of your albums are there (type albums into the R Console to see your list of albums after exiting the comparison loop).
It ended up saving, I had to "unzip" the file. However, I am now running into a new problem. I cannot view my list.


I should have explained that you need to unzip the folder!

In terms of viewing the list, do you mean within R? Are you able to get into the comparison loop? This loop will continue until you press something other than 1 or 2. Once you exit the comparison loop, you can see your ranked list by typing display and hitting Enter.
When I type display into R, after exiting the loop, it says Error: object 'display' not found.

Re: The list-making workshop : sharing our list-making methods

Listyguy
Moonbeam
Listyguy
Moonbeam
Are you editing your album list using Microsoft Excel, or some other software?

I use Excel, and it always prompts me "Album List.txt may contain features that are not compatible with Text (Tab delimited)."

It never seems to be a problem, though. As long as the program is working and all of your albums are there (type albums into the R Console to see your list of albums after exiting the comparison loop).
It ended up saving, I had to "unzip" the file. However, I am now running into a new problem. I cannot view my list.


I should have explained that you need to unzip the folder!

In terms of viewing the list, do you mean within R? Are you able to get into the comparison loop? This loop will continue until you press something other than 1 or 2. Once you exit the comparison loop, you can see your ranked list by typing display and hitting Enter.
When I type display into R, after exiting the loop, it says Error: object 'display' not found.


Is anybody else experiencing this problem? It's strange, because display is redefined every time you exit the loop.

Would you mind telling me what comes up when you type Rating?

The only thing I can think of is that perhaps you didn't install the KernSmooth package, because it is used to calculate Rating, and that is done before display is defined.

If it also returns an error when you type Rating, it indicates that either you haven't installed the KernSmooth package or it isn't loading properly.

Re: The list-making workshop : sharing our list-making methods

Moonbeam- Not sure if you can help with this, but everytime I click on the nearest location to install the KernSmooth package, I receive this error message:
> utils:::menuInstallPkgs()
--- Please select a CRAN mirror for use in this session ---
Warning: unable to access index for repository http://lib.stat.cmu.edu/R/CRAN/bin/windows/contrib/2.13
Warning: unable to access index for repository http://www.stats.ox.ac.uk/pub/RWin/bin/windows/contrib/2.13
Error in install.packages(NULL, .libPaths()[1L], dependencies = NA, type = type) :
no packages were specified
In addition: Warning message:
In open.connection(con, "r") : cannot open: HTTP status was '0 (nil)'

Any ideas as to how to fix this?

Thanks!

Re: The list-making workshop : sharing our list-making methods

I had the "display cannot be found" error and it turned out that my album list file wasn't right.
So if you have this error check that your album lists txt file is right, with tabulations and columns at the right place.

Re: The list-making workshop : sharing our list-making methods

Moonbeam, thanks for the answer regarding the stars.

And everything works fine for me, so no display errors or so ...

The only disadvantage of the tool is, that you need to run the music near or on the PC. It would be nice if the code allows for reading text files with results of matchups (this could also be tab delimited with "ID1 ID2 winner" per line). I'm able to understand the major part of the code by now and it should be able to add the necessary code for reading such files... Moonbeam, did you already program something into this direction?

Re: The list-making workshop : sharing our list-making methods

You can disregard my error message, as I did get this to work... pretty damn cool!

Re: The list-making workshop : sharing our list-making methods

Awesome program Moonbeam. It's going to take awhile to get a good list since I exported my RYM data into it but in a couple weeks it's probably going to be the most accurate list I've come up with. Thanks so much!

Re: The list-making workshop : sharing our list-making methods

Andre
Moonbeam, thanks for the answer regarding the stars.

And everything works fine for me, so no display errors or so ...

The only disadvantage of the tool is, that you need to run the music near or on the PC. It would be nice if the code allows for reading text files with results of matchups (this could also be tab delimited with "ID1 ID2 winner" per line). I'm able to understand the major part of the code by now and it should be able to add the necessary code for reading such files... Moonbeam, did you already program something into this direction?


Andre, I don't quite understand how having a text file for matchup results would allow you to run the program away from a PC or laptop. You'll still need to access R to run it. Would you mind expounding?

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Is anybody else experiencing this problem? It's strange, because display is redefined every time you exit the loop.

Would you mind telling me what comes up when you type Rating?

The only thing I can think of is that perhaps you didn't install the KernSmooth package, because it is used to calculate Rating, and that is done before display is defined.

If it also returns an error when you type Rating, it indicates that either you haven't installed the KernSmooth package or it isn't loading properly.
When I type rating, it retuns the message Error: object 'Rating' not found. I don't think the KernSmooth is the problem, because when I tried to reinstall it, R told me I couldn't because it was in use.
I don't know if this helps you any, but when I try to use the artist function, I get the return message Error: could not find function "artist".

Edit: Could this be because I don't have star ratings for the albums?

Re: The list-making workshop : sharing our list-making methods

Listyguy
Moonbeam
Is anybody else experiencing this problem? It's strange, because display is redefined every time you exit the loop.

Would you mind telling me what comes up when you type Rating?

The only thing I can think of is that perhaps you didn't install the KernSmooth package, because it is used to calculate Rating, and that is done before display is defined.

If it also returns an error when you type Rating, it indicates that either you haven't installed the KernSmooth package or it isn't loading properly.
When I type rating, it retuns the message Error: object 'Rating' not found. I don't think the KernSmooth is the problem, because when I tried to reinstall it, R told me I couldn't because it was in use.
I don't know if this helps you any, but when I try to use the artist function, I get the return message Error: could not find function "artist".

Edit: Could this be because I don't have star ratings for the albums?


I'm pretty positive that it is because you don't have star ratings. I should have programmed it so that the Rating column was optional, and perhaps I'll do that and share another version with the forum. If you don't care much about the Rating, just enter two ratings that are different somewhere in the spreadsheet and it should be ok.

Edit: Crap, I've just found that star ratings need to be entered for all albums for it to be ok. I'll work on a rewrite and hopefully get it up later today.

Edit 2: Listyguy, you can get the program to work if you enter just 2 different star ratings if you edit the 'Album Ranker.txt' file. If you can open it with a Text viewing program (like Notepad), go to the line immediately under library(KernSmooth) that reads:

star.rating = 20*stars

Replace this line with the following line:

star.rating = 20*stars[is.na(stars) == FALSE]

Save the file. This should provide a quick fix, but I'll still look into making another version of the program without needing to enter any Star ratings.

Re: The list-making workshop : sharing our list-making methods

I keep my favorite SONGS in folders: 0-100 and so on. Occasionally some songs get the call-up or gets demoted. New favorites start at the bottom. Like in working life.

Albums: "Closer" is much better than the sum of it's parts. "Marquee Moon" doesn't have a weak spot. "Cosmos Factory" has good songs but is relatively uninterresting. I guess albums must be rated in some kind of complex way that has too many factors to be counted. I guess I use my "good-bad" neural brain centre. I don't believe in apply ratings for each song on an album.

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Andre
Moonbeam, thanks for the answer regarding the stars.

And everything works fine for me, so no display errors or so ...

The only disadvantage of the tool is, that you need to run the music near or on the PC. It would be nice if the code allows for reading text files with results of matchups (this could also be tab delimited with "ID1 ID2 winner" per line). I'm able to understand the major part of the code by now and it should be able to add the necessary code for reading such files... Moonbeam, did you already program something into this direction?


Andre, I don't quite understand how having a text file for matchup results would allow you to run the program away from a PC or laptop. You'll still need to access R to run it. Would you mind expounding?


What I meant is something like:
1) listen to songs on an arbitrary place (not necessary near a PC)
2) at your PC, enter matchup results in text file
3) run R to read and integrate text files

This would be convenient for me, because then I can listen to a preselected list of songs during the day and enter the results electronically in the evening.

Going to bed now ... this program of yours might keep me from sleeping though

Re: The list-making workshop : sharing our list-making methods

I'm loving the matchup aspect. Sometimes I think that you programmed to come up with fun matchups but it's my list so that's impossible right? One matchup this morning was Blonde Redhead- 23 vs. Adele- 21.

Re: The list-making workshop : sharing our list-making methods

Andre
Moonbeam
Andre
Moonbeam, thanks for the answer regarding the stars.

And everything works fine for me, so no display errors or so ...

The only disadvantage of the tool is, that you need to run the music near or on the PC. It would be nice if the code allows for reading text files with results of matchups (this could also be tab delimited with "ID1 ID2 winner" per line). I'm able to understand the major part of the code by now and it should be able to add the necessary code for reading such files... Moonbeam, did you already program something into this direction?


Andre, I don't quite understand how having a text file for matchup results would allow you to run the program away from a PC or laptop. You'll still need to access R to run it. Would you mind expounding?


What I meant is something like:
1) listen to songs on an arbitrary place (not necessary near a PC)
2) at your PC, enter matchup results in text file
3) run R to read and integrate text files

This would be convenient for me, because then I can listen to a preselected list of songs during the day and enter the results electronically in the evening.

Going to bed now ... this program of yours might keep me from sleeping though


So you want to control the matchups based on recent listens? That's not a bad idea, but I suppose it's a different approach to the random matchup scheme. I'll see what I can come up with!

Re: The list-making workshop : sharing our list-making methods

John
I'm loving the matchup aspect. Sometimes I think that you programmed to come up with fun matchups but it's my list so that's impossible right? One matchup this morning was Blonde Redhead- 23 vs. Adele- 21.


It must be coincidence. I enjoy my matchups that pit IV by Gap Band against IV by TOTO!

Re: The list-making workshop : sharing our list-making methods

Listyguy (and everyone), I'll have an updated version of the program posted within a few hours.

Would anybody be interested in a song version of this? It wouldn't be difficult to adapt one from the current album program.

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Edit 2: Listyguy, you can get the program to work if you enter just 2 different star ratings if you edit the 'Album Ranker.txt' file. If you can open it with a Text viewing program (like Notepad), go to the line immediately under library(KernSmooth) that reads:

star.rating = 20*stars

Replace this line with the following line:

star.rating = 20*stars[is.na(stars) == FALSE]

Save the file. This should provide a quick fix, but I'll still look into making another version of the program without needing to enter any Star ratings.
I edited the "Album Ranker" file, but now what happens is when I exit the loop, the R retunrs this message:
KernSmooth 2.23 loaded
Copyright M. P. Wand 1997-2009
Error in if (L == 0) warning("Binning grid too coarse for current (small) bandwidth: consider increasing 'gridsize'") :
missing value where TRUE/FALSE needed

I gave two albums ratings (One a 5 and the other a 3), if that's any help.
Maybe I should just wait for the other version.

Re: The list-making workshop : sharing our list-making methods

Here is the link to the updated version of the program.

Listyguy, I can see that the kernel density estimation could have problems when there are few star ratings relative to the number of albums in the database. Do you have a big list of albums? I've tested it now with only 2 star ratings (5 and 3) out of a list of 800+ albums. It took awhile to find a density (about 10 seconds), but it did it.

My advice would be to enter more ratings as you see fit. If you really don't care about the Rating part, this new version of the software should work if you delete all of your entries in the Star column.

Re: The list-making workshop : sharing our list-making methods

No spreadsheets, no numbers, no programs, no anything.

When I made my first list of top 200 albums and top 500 songs, I went through my iTunes playlist and when I saw something that I like, I added the album/song to a .doc list. Both of these lists ended up being around twice as large as what the list needs to be, so I start cutting until I end up with the something close to or even less than the final list size. Then I signal out my albums/songs that I consider my "canon" and figure out which album I would put as the "gatekeeper" for each level (for example, The Replacement's Let It Be is my canon gatekeeper for albums, while Weird Al's Albuquerque is that for songs.)

Once I establish the gatekeeper I start looking through the list for things I consider better than that, and then work out an order one-by-one. Takes alot of moving around to get there. And even then, I look at my lists from 2 years ago and wonder how I could think that x is better than y. I'm working on my top 200 albums again, I put it away for a month and need to do a long look-over for everything again. I've taken my original list and added around 25-30 albums that I would consider adding to the list: some will make it, some won't.

Re: The list-making workshop : sharing our list-making methods

This is more addictive than crack Moonbeam.

Re: The list-making workshop : sharing our list-making methods

1,1,2,1,2,2,1,2,1,2,1,2,1,2,1,2,1,2,2,1,2,1,2,1,2,1,2,1,2,1,2,2,1,1,2,1,2,2,2,1

Re: The list-making workshop : sharing our list-making methods

I'm glad you guys are enjoying it!

Re: The list-making workshop : sharing our list-making methods

Here's a couple questions Ian:

How do you know when you have a good list? Just from spending a few hours it seems like the program is trying to guess what I'd like more by putting the perceived best album as #1. Am I correct? If that's the case, I would assume that once it comes to the point where you are picking #1 all the time, the list is complete.

Also, is it possible to add albums to the Album List once I've started ranking?

Re: The list-making workshop : sharing our list-making methods

John
Also, is it possible to add albums to the Album List once I've started ranking?
I added a second list to the first after a day of ranking and it blended seamlessly, so yes to your second question.

Re: The list-making workshop : sharing our list-making methods

Moonbeam

So you want to control the matchups based on recent listens? That's not a bad idea, but I suppose it's a different approach to the random matchup scheme. I'll see what I can come up with!


Exactly and thanks in advance.

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Listyguy (and everyone), I'll have an updated version of the program posted within a few hours.

Would anybody be interested in a song version of this? It wouldn't be difficult to adapt one from the current album program.


Moonbeam, I'm already using it for songs. What would be different? The column title "Album" in the report doesn't bother me.

But to answer your question: Yes, I would be interested in a songs version.

Re: The list-making workshop : sharing our list-making methods

John
Here's a couple questions Ian:

How do you know when you have a good list? Just from spending a few hours it seems like the program is trying to guess what I'd like more by putting the perceived best album as #1. Am I correct? If that's the case, I would assume that once it comes to the point where you are picking #1 all the time, the list is complete.

Also, is it possible to add albums to the Album List once I've started ranking?


John, this is a difficult question. Do you mean #1 as in the first album that appears in a random comparison? That is not the case. If you retain the set = "balance" setting, the program will cycle through all of the albums in order of the least amount of comparisons.

That said, the program will narrow its focus for each comparison as the point distribution spreads. I found that for my 732 album database, it identified my #1 album after each album had only 10 comparisons, and this was quicker than either of the other 2 point systems I had tried. The top and bottom of the list should align with your true list first, and then the middle. As for when it is complete, I can't really answer that.

You can check how well it has converged at the top and bottom by doing this:

Type sum(winpct == 100) into the Console. This will tell you how many albums are currently undefeated.

Type sum(winpct == 0) into the Console. This will tell you how many albums are currently winless.

These should eventually turn out to be 1.

Re: The list-making workshop : sharing our list-making methods

Brad
John
Also, is it possible to add albums to the Album List once I've started ranking?
I added a second list to the first after a day of ranking and it blended seamlessly, so yes to your second question.


This is correct. As you hear new albums, you can simply add them to the 'Album List.txt' file. What's great about keeping set = "balance" is that the program will continue to include these new entries in every step of the comparison loop until they catch up to the minimum number of comparisons of the older entries.

Re: The list-making workshop : sharing our list-making methods

Andre
Moonbeam
Listyguy (and everyone), I'll have an updated version of the program posted within a few hours.

Would anybody be interested in a song version of this? It wouldn't be difficult to adapt one from the current album program.


Moonbeam, I'm already using it for songs. What would be different? The column title "Album" in the report doesn't bother me.

But to answer your question: Yes, I would be interested in a songs version.


In principle, nothing is different except a few labels. The program would have to read in the appropriate Song text file and save the relevant vectors in a Song RData file so as to not be confused with the Album files. That's really all.

Re: The list-making workshop : sharing our list-making methods

When you say M * M, you're talking about matrix multiplication, right? That's an interesting approach. Every entry of M * M becomes a measure of whether the column album has lost to an album the row album has beaten. The effect is, every time an album gets a point, every album that beat it also gets a point.

This seems like a very good snapshot of your current taste, but it doesn't seem sensitive to changing tastes. In my algorithm, the weight of any given result is gradually reduced, which happens faster for newly acquired albums than albums I've had for a while, and a new match between two albums erases the result of the previous match.

Re: The list-making workshop : sharing our list-making methods

BillAdama
When you say M * M, you're talking about matrix multiplication, right? That's an interesting approach. Every entry of M * M becomes a measure of whether the column album has lost to an album the row album has beaten. The effect is, every time an album gets a point, every album that beat it also gets a point.

This seems like a very good snapshot of your current taste, but it doesn't seem sensitive to changing tastes. In my algorithm, the weight of any given result is gradually reduced, which happens faster for newly acquired albums than albums I've had for a while, and a new match between two albums erases the result of the previous match.


Indeed! The program actually multiplies the matchup result matrix a number of times, so that not only does each album reflect its opponents and opponents' opponents, but several more levels. I suppose I could introduce a weight matrix to include as well, but the program has worked well enough for me to be reasonably happy with it.

Re: The list-making workshop : sharing our list-making methods

Moonbeam, just wanted to let you know, I decided to give my albums ratings to make everything easier, and the program is working well. However, when I try to use the other functions (year, artist), I get the return message:

R
Error: could not find function "artist"

Re: The list-making workshop : sharing our list-making methods

Moonbeam
John
Here's a couple questions Ian:

How do you know when you have a good list? Just from spending a few hours it seems like the program is trying to guess what I'd like more by putting the perceived best album as #1. Am I correct? If that's the case, I would assume that once it comes to the point where you are picking #1 all the time, the list is complete.

Also, is it possible to add albums to the Album List once I've started ranking?


John, this is a difficult question. Do you mean #1 as in the first album that appears in a random comparison? That is not the case. If you retain the set = "balance" setting, the program will cycle through all of the albums in order of the least amount of comparisons.

That said, the program will narrow its focus for each comparison as the point distribution spreads. I found that for my 732 album database, it identified my #1 album after each album had only 10 comparisons, and this was quicker than either of the other 2 point systems I had tried. The top and bottom of the list should align with your true list first, and then the middle. As for when it is complete, I can't really answer that.

You can check how well it has converged at the top and bottom by doing this:

Type sum(winpct == 100) into the Console. This will tell you how many albums are currently undefeated.

Type sum(winpct == 0) into the Console. This will tell you how many albums are currently winless.

These should eventually turn out to be 1.


OK. It just felt like I was picking mostly 1's. But, my list is starting to look really good. Better, than I could have done on paper already.

What if I replaced Album List.txt with a new one? If everything that was on the old list was still named the same would it change anything? This is for the purpose of continuing to just export RYM data.... although I don't know how much I'll be using RYM as my main database now that I have this.

Re: The list-making workshop : sharing our list-making methods

Listyguy
Moonbeam, just wanted to let you know, I decided to give my albums ratings to make everything easier, and the program is working well. However, when I try to use the other functions (year, artist), I get the return message:
R
Error: could not find function "artist"


That's strange. The year and artist functions are defined immediately after display, so if that works, those functions should work. Perhaps try exiting the program completely and going in again, so that the entire script can be run.

Re: The list-making workshop : sharing our list-making methods

Moonbeam
BillAdama
When you say M * M, you're talking about matrix multiplication, right? That's an interesting approach. Every entry of M * M becomes a measure of whether the column album has lost to an album the row album has beaten. The effect is, every time an album gets a point, every album that beat it also gets a point.

This seems like a very good snapshot of your current taste, but it doesn't seem sensitive to changing tastes. In my algorithm, the weight of any given result is gradually reduced, which happens faster for newly acquired albums than albums I've had for a while, and a new match between two albums erases the result of the previous match.


Indeed! The program actually multiplies the matchup result matrix a number of times, so that not only does each album reflect its opponents and opponents' opponents, but several more levels. I suppose I could introduce a weight matrix to include as well, but the program has worked well enough for me to be reasonably happy with it.


Oh, in that case, it mathematically works out very similar to the way mine does. My original idea was to use the wins to make a directed graph and then count the maximum number of steps it took to reach each album, but that would have had to check for cycles every time a win was added and just delete the oldest edge.

I wonder, if I used this algorithm to build a commercial website similar to Criticker (Which would allow you to search out people with taste similar to yours, and build a recommendation list based on this formula), could I make any money off it?

Even if I didn't make any money it'd make a fantastic resume line...

Re: The list-making workshop : sharing our list-making methods

John
Moonbeam
John
Here's a couple questions Ian:

How do you know when you have a good list? Just from spending a few hours it seems like the program is trying to guess what I'd like more by putting the perceived best album as #1. Am I correct? If that's the case, I would assume that once it comes to the point where you are picking #1 all the time, the list is complete.

Also, is it possible to add albums to the Album List once I've started ranking?


John, this is a difficult question. Do you mean #1 as in the first album that appears in a random comparison? That is not the case. If you retain the set = "balance" setting, the program will cycle through all of the albums in order of the least amount of comparisons.

That said, the program will narrow its focus for each comparison as the point distribution spreads. I found that for my 732 album database, it identified my #1 album after each album had only 10 comparisons, and this was quicker than either of the other 2 point systems I had tried. The top and bottom of the list should align with your true list first, and then the middle. As for when it is complete, I can't really answer that.

You can check how well it has converged at the top and bottom by doing this:

Type sum(winpct == 100) into the Console. This will tell you how many albums are currently undefeated.

Type sum(winpct == 0) into the Console. This will tell you how many albums are currently winless.

These should eventually turn out to be 1.


OK. It just felt like I was picking mostly 1's. But, my list is starting to look really good. Better, than I could have done on paper already.

What if I replaced Album List.txt with a new one? If everything that was on the old list was still named the same would it change anything? This is for the purpose of continuing to just export RYM data.... although I don't know how much I'll be using RYM as my main database now that I have this.


The problem with doing that is that RYM will place newer ratings in the list alphabetically, while the order matters in text file, because it determines the relevant row and column of the results matrix. For example, say your favorite album is currently 93rd in the text file, but will become 95th once you export from RYM with the newer albums. All of the hard work your #1 album put in to get to #1 will then be credited to whatever the 93rd album in the new text file would be!

The only approach for retaining the correct results matrix is unfortunately to add new albums to the bottom of the text file.

Re: The list-making workshop : sharing our list-making methods

BillAdama
Moonbeam
BillAdama
When you say M * M, you're talking about matrix multiplication, right? That's an interesting approach. Every entry of M * M becomes a measure of whether the column album has lost to an album the row album has beaten. The effect is, every time an album gets a point, every album that beat it also gets a point.

This seems like a very good snapshot of your current taste, but it doesn't seem sensitive to changing tastes. In my algorithm, the weight of any given result is gradually reduced, which happens faster for newly acquired albums than albums I've had for a while, and a new match between two albums erases the result of the previous match.


Indeed! The program actually multiplies the matchup result matrix a number of times, so that not only does each album reflect its opponents and opponents' opponents, but several more levels. I suppose I could introduce a weight matrix to include as well, but the program has worked well enough for me to be reasonably happy with it.


Oh, in that case, it mathematically works out very similar to the way mine does. My original idea was to use the wins to make a directed graph and then count the maximum number of steps it took to reach each album, but that would have had to check for cycles every time a win was added and just delete the oldest edge.

I wonder, if I used this algorithm to build a commercial website similar to Criticker (Which would allow you to search out people with taste similar to yours, and build a recommendation list based on this formula), could I make any money off it?

Even if I didn't make any money it'd make a fantastic resume line...


I'm not sure how lucrative it would be, as I don't think the demand for music recommendations is the same as it is for films, but it would be interesting!

Re: The list-making workshop : sharing our list-making methods

Well, I was thinking of having functionality for both music and films. I also wonder if I could have some kind of wiki-style album description section.

Also, I think there is a latent demand for music recommendations among those who want to 'Beat the crowd' to the next big thing.

I'm not hoping for something that would pay the bills, so much as something that could give me enough pocket money so I can add a second entry to my 'Work history' section.

Re: The list-making workshop : sharing our list-making methods

Moonbeam, I'd be intersted in a song ranker, and an artist one as well;
I already started a song ranking with your system like Andre did.
I installed your program on both my office and home computers
My question is : can I run two album rankers on the same computer or do i have to change the files' names ?

thanx in advance

Re: The list-making workshop : sharing our list-making methods

nicolas
Moonbeam, I'd be intersted in a song ranker, and an artist one as well;
I already started a song ranking with your system like Andre did.
I installed your program on both my office and home computers
My question is : can I run two album rankers on the same computer or do i have to change the files' names ?

thanx in advance


On my PC, I have two workspaces, one for the albums (Albums.RData) and one for songs (Songs.RData). But you could also have two workspaces for albums. So I think the answer to your question is, yes you can run two album rankers on the same computer by using two different workspaces. Moonbeam, can you confirm?

Re: The list-making workshop : sharing our list-making methods

nicolas
Moonbeam, I'd be intersted in a song ranker, and an artist one as well;
I already started a song ranking with your system like Andre did.
I installed your program on both my office and home computers
My question is : can I run two album rankers on the same computer or do i have to change the files' names ?

thanx in advance


You can run two album rankers on the same computer as long as they are in different folders. If one of them is for ranking songs, I'll upload a version for songs alone soon.

Re: The list-making workshop : sharing our list-making methods

Andre
nicolas
Moonbeam, I'd be intersted in a song ranker, and an artist one as well;
I already started a song ranking with your system like Andre did.
I installed your program on both my office and home computers
My question is : can I run two album rankers on the same computer or do i have to change the files' names ?

thanx in advance


On my PC, I have two workspaces, one for the albums (Albums.RData) and one for songs (Songs.RData). But you could also have two workspaces for albums. So I think the answer to your question is, yes you can run two album rankers on the same computer by using two different workspaces. Moonbeam, can you confirm?


Correct!

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Brad
John
Also, is it possible to add albums to the Album List once I've started ranking?
I added a second list to the first after a day of ranking and it blended seamlessly, so yes to your second question.


This is correct. As you hear new albums, you can simply add them to the 'Album List.txt' file. What's great about keeping set = "balance" is that the program will continue to include these new entries in every step of the comparison loop until they catch up to the minimum number of comparisons of the older entries.


So, if I add an album it will automatically give me 10 different situations involving that album if I've ranked every other album 10 times?

Re: The list-making workshop : sharing our list-making methods

John
Moonbeam
Brad
John
Also, is it possible to add albums to the Album List once I've started ranking?
I added a second list to the first after a day of ranking and it blended seamlessly, so yes to your second question.


This is correct. As you hear new albums, you can simply add them to the 'Album List.txt' file. What's great about keeping set = "balance" is that the program will continue to include these new entries in every step of the comparison loop until they catch up to the minimum number of comparisons of the older entries.


So, if I add an album it will automatically give me 10 different situations involving that album if I've ranked every other album 10 times?


Yes, if set = "balance".

Re: The list-making workshop : sharing our list-making methods

Fred
"Cosmos Factory" has good songs but is relatively uninterresting.


To me this sentence makes no sense, and not only because I love Cosmos Factory

If the album has good songs then it's good. Period. "Interesting" is not a criteria to me. Too intellectual. I use a heart factor. If the songs fail to involve me personnaly then they're no good and the album's not a favorite. Of course it all depends on our own definition of "good" and "intersting".


I just think the other way round. For example, "69 Love Songs" is interesting (different, conceptual) but has not enough good songs. So : out of my list

"Doolittle" has songs that I might consider good if I liked the Pixies' style (typical exemple : the song with the countdown, sorry i can't remember its name) ; but I don't like their sometimes too aggressive and twisted music . So : out (even if I like some of the Pixies songs, like "Here Comes Your Man")

IMO there's a difference between "good" and "well made". "Since I Left You" by the Avalanches is well made but fails to resonate emotionally. Out.

Re: The list-making workshop : sharing our list-making methods

For me 'Interesting' doesn't imply intellectual interest when applied to music. When I call music 'Interesting' I mean 'It breaks my expectations'. If an album sounds good, but sounds exactly the same as something better, I will rate it a lot lower than something that sounds equally good but is unlike anything else.

Also, when I say an album 'Has good songs' I generally mean 'Has a few good songs', with the implication that the album isn't consistent. Maybe that's what he meant in reference to Cosmo's Factory.

Re: The list-making workshop : sharing our list-making methods

OK, I added an album to the bottom of my Album List.txt and now it's omitting the artist name from every choice and when I stop the cycle I get an error. Any ideas what I might have done wrong?

Re: The list-making workshop : sharing our list-making methods

John
OK, I added an album to the bottom of my Album List.txt and now it's omitting the artist name from every choice and when I stop the cycle I get an error. Any ideas what I might have done wrong?


Hmmm. Did you enter an album and an artist field? Would you mind telling me what they are? R can have problems with certain symbols like apostrophes and things, but this shouldn't prevent it from working. Alternatively, you could email me the text file and I can fiddle with it.

Re: The list-making workshop : sharing our list-making methods

I'll take a look at it when I get home and give you the text file if I can't figure it out.

Re: The list-making workshop : sharing our list-making methods

John
I'll take a look at it when I get home and give you the text file if I can't figure it out.


I hope it works out. The error suggests to me that there is an uneven number of rows for each column. Text files are strange beasts in that if you delete something, it still retains a placeholder in that cell.

Re: The list-making workshop : sharing our list-making methods

Andre
Moonbeam

So you want to control the matchups based on recent listens? That's not a bad idea, but I suppose it's a different approach to the random matchup scheme. I'll see what I can come up with!


Exactly and thanks in advance.


Quick question, Andre. How would you want to enter in the information for Album 1 and Album 2 for each direct comparison? As their numeric place in the list, or as "Artist 1 - Album 1" and "Artist 2 - Album 2"?

Re: The list-making workshop : sharing our list-making methods

Andre
Moonbeam, thanks for the answer regarding the stars.

And everything works fine for me, so no display errors or so ...

The only disadvantage of the tool is, that you need to run the music near or on the PC. It would be nice if the code allows for reading text files with results of matchups (this could also be tab delimited with "ID1 ID2 winner" per line). I'm able to understand the major part of the code by now and it should be able to add the necessary code for reading such files... Moonbeam, did you already program something into this direction?


Moonbeam, my suggestion would be to use the IDs. Each album/song has an ID, so the input file could look like:

ID1 ID2 winner
23 46 1
25 132 2

etcetera. The file is tab delimited, so you can use the same read.delim command in your code.

In this case you need to lookup the IDs, but entering album/artist names or the current ranking position is more error prone, I would say.

Re: The list-making workshop : sharing our list-making methods

Andre
Andre
Moonbeam, thanks for the answer regarding the stars.

And everything works fine for me, so no display errors or so ...

The only disadvantage of the tool is, that you need to run the music near or on the PC. It would be nice if the code allows for reading text files with results of matchups (this could also be tab delimited with "ID1 ID2 winner" per line). I'm able to understand the major part of the code by now and it should be able to add the necessary code for reading such files... Moonbeam, did you already program something into this direction?


Moonbeam, my suggestion would be to use the IDs. Each album/song has an ID, so the input file could look like:

ID1 ID2 winner
23 46 1
25 132 2

etcetera. The file is tab delimited, so you can use the same read.delim command in your code.

In this case you need to lookup the IDs, but entering album/artist names or the current ranking position is more error prone, I would say.


That should be easy enough to implement! I'll try to get this up in addition to a Songs version within the next 3 days or so.

Would people also be interested in an artist ranking program based on Henrik's formula? The only thing we'd have to work out is how to tweak it for different sized lists. Perhaps Henrik can help here. Also, I suppose we'd also have to have an optimized cutoff for inclusion in the formula, as some people may include albums/songs in their list that they don't want to count toward artist ranking, particularly if they include stuff they don't like!

Re: The list-making workshop : sharing our list-making methods

So I've decided to take the plunge and work a little bit with Moonbeam's program. It reminds me of the website Flickchart, but that is for movies. There's no way to add albums to your list en masse right? It may take a while to add my whole collection one at a time.

Re: The list-making workshop : sharing our list-making methods

Vanillafire1000
So I've decided to take the plunge and work a little bit with Moonbeam's program. It reminds me of the website Flickchart, but that is for movies. There's no way to add albums to your list en masse right? It may take a while to add my whole collection one at a time.
IF you have your list in excell, you can copy and paste it directly from there.

Re: The list-making workshop : sharing our list-making methods

Vanillafire1000
So I've decided to take the plunge and work a little bit with Moonbeam's program. It reminds me of the website Flickchart, but that is for movies. There's no way to add albums to your list en masse right? It may take a while to add my whole collection one at a time.


Is there any way that you can export your album list from wherever you have it stored into a text file of some sort? You could then just copy and paste the relevant columns.

Re: The list-making workshop : sharing our list-making methods

I'm going to try to make a new version of mine that allows you to add in albums by entering them into tab delimited files, so you can add them in straight through Excel.

I will also make it flexible as far as fields, and allow you to filter by fields. So, you could do it with music and use year, band, etc. Or, do it with film and use year, director, writer, etc. Or, do it with sports teams, etc. And make it possible to customize the weighting.

Re: The list-making workshop : sharing our list-making methods

Moonbeam
Vanillafire1000
So I've decided to take the plunge and work a little bit with Moonbeam's program. It reminds me of the website Flickchart, but that is for movies. There's no way to add albums to your list en masse right? It may take a while to add my whole collection one at a time.


Is there any way that you can export your album list from wherever you have it stored into a text file of some sort? You could then just copy and paste the relevant columns.


None of it comes out formatted properly (using iTunes as a library and all.) I could probably get it to work if I knew anything about spreadsheets, but its too much of a hassle for me. I've done data entry before, I can do this monotonous work.

Re: The list-making workshop : sharing our list-making methods

Okay, update, I'm going to try to make this program by the end of the week because a recruiting firm I'm working with is asking for code samples.

So, I want to make a description of the program to work with, to see if anybody has ideas or suggestions.

At this stage it will still run on the Java command line, but I want to make it so you can import information from a CSV file which has been created from an Excel file.

So, here's how it will work. The top row of the Excel file will be parameters. You can add anything you want, but it will come with three templates, one for music albums, one for music songs, and one for films. Additionally you can make the parameters however you want.

You would initiate the program with a CSV file and it would load it in. Then, it would give you a list of options:
1) File menu
-1) Load new file
-2) Save
-3) Add parameter
2) Data menu
-1) Add result (This would allow you to list any number of items, best to worst)
-2) Delete result
-3) Print results (This would prompt you with a filter menu to ask which albums' results you want to print)
3) Score menu
-1) Add filter
-2) Remove all filters
(These would allow you to only list albums from the same year, the same artist, or whatever)
-3) Print scores
-4) Update scores
-5) Recommend next match
4) Preferences menu

The preferences menu would allow you to fiddle with things. For instance, you could decide whether every match is equal weight or if weights gradually decline. You could decide whether new results add to or overwrite existing results.

Comments? I want to have this done and distributable really quickly.

Re: The list-making workshop : sharing our list-making methods

Time for a short anecdote related to all this list-mania:

At the end of the 90s I used to work at a small company that develops systems for behavioral research. Based on an existing DOS program called MatMan (short for matrix manipulation), I developed in about 6 months an Excel add-in which had completely ported the DOS version and had some functionality extended to it (something complicated with the statistical correlation parameters Pearson's R and Spearman's rho).

One of the program functions is to retrieve a hierarchy and ranking within a group of animals. If you, for instance, observe a group of monkeys for a while and take a note each time two monkeys get agressive, or like each other and start fleeing each other, you can enter this data in Excel and run the program. It will give you the hierarchy.

I still should have the VBA code somewhere, maybe it will work for music albums and songs as well?

And I'm happy to see that the program is still being sold: http://www.noldus.com/human-behavior-research/products/matman

Re: The list-making workshop : sharing our list-making methods

BillAdama
Okay, update, I'm going to try to make this program by the end of the week because a recruiting firm I'm working with is asking for code samples.

So, I want to make a description of the program to work with, to see if anybody has ideas or suggestions.

At this stage it will still run on the Java command line, but I want to make it so you can import information from a CSV file which has been created from an Excel file.

So, here's how it will work. The top row of the Excel file will be parameters. You can add anything you want, but it will come with three templates, one for music albums, one for music songs, and one for films. Additionally you can make the parameters however you want.

You would initiate the program with a CSV file and it would load it in. Then, it would give you a list of options:
1) File menu
-1) Load new file
-2) Save
-3) Add parameter
2) Data menu
-1) Add result (This would allow you to list any number of items, best to worst)
-2) Delete result
-3) Print results (This would prompt you with a filter menu to ask which albums' results you want to print)
3) Score menu
-1) Add filter
-2) Remove all filters
(These would allow you to only list albums from the same year, the same artist, or whatever)
-3) Print scores
-4) Update scores
-5) Recommend next match
4) Preferences menu

The preferences menu would allow you to fiddle with things. For instance, you could decide whether every match is equal weight or if weights gradually decline. You could decide whether new results add to or overwrite existing results.

Comments? I want to have this done and distributable really quickly.


This all sounds amazing! The only thing I'd add is the capacity to present lists by decade (or perhaps more generally any user-defined period).

Re: The list-making workshop : sharing our list-making methods

Andre
Time for a short anecdote related to all this list-mania:

At the end of the 90s I used to work at a small company that develops systems for behavioral research. Based on an existing DOS program called MatMan (short for matrix manipulation), I developed in about 6 months an Excel add-in which had completely ported the DOS version and had some functionality extended to it (something complicated with the statistical correlation parameters Pearson's R and Spearman's rho).

One of the program functions is to retrieve a hierarchy and ranking within a group of animals. If you, for instance, observe a group of monkeys for a while and take a note each time two monkeys get agressive, or like each other and start fleeing each other, you can enter this data in Excel and run the program. It will give you the hierarchy.

I still should have the VBA code somewhere, maybe it will work for music albums and songs as well?

And I'm happy to see that the program is still being sold: http://www.noldus.com/human-behavior-research/products/matman


That's awesome, Andre! I'm off to check it out.

Re: The list-making workshop : sharing our list-making methods

Okay, how about this: Template files specify whether it takes integer arguments or string arguments, and if it takes integer arguments, you can specify a whole range.

Re: The list-making workshop : sharing our list-making methods

BillAdama
Okay, how about this: Template files specify whether it takes integer arguments or string arguments, and if it takes integer arguments, you can specify a whole range.


1 2