This might also be a \"feature request\" but it\'s more a bone to pick with the way score is handled.

I would like a way to skip, or at least stop, a song without affecting its score. Sometimes I\'ll have a generic playlist of top-rated songs but I might not be in the mood for a particular song, ie it\'s a great song but a bit depressing and I just don\'t want to listen to it at that second, or I\'m turning in for the night and want to stop the player but without affecting score. With the way it is now I have to open the playlist, remember the song, skip or stop it, go to the next one, and then go back and change its score back. I\'m constantly worried about affecting the score when I don\'t want to. (The easiest way to deal with that is make it so just stopping a song doesn\'t call the score algorithm. Users who don\'t want to affect score but do want to skip songs could stop then next and play)

I also feel that once a song has played at least half way through its score shouldn\'t go down. Most people that don\'t like a song will have skipped it within the first half and if you\'ve listened to at least half the song you obviously like it to one degree or another. I\'d recommend making the halfway point a \'no score change\' and anything after that would increase the score as if it played to the end, except it would get less points than if it was played all the way.

I know the manual for score says it won\'t change the score but by a few points, however that is utter bull. I have a relatively new collection of songs and most songs have been played less than ten to fifteen times. Any time I skip a song the score usually changes by over ten points. In order to keep the score from changing so much the \"+ 1\" should be increased to something like \"+10\" or \"+15\", or add a \"if played < 10/15 played = 10/15\". If someone really doesn\'t like a song that much it\'s easier to manually lower the score after the first play than to keep raising after each stop.

I\'ve made multiple smart playlists just to deal with these issues.
Score > 50, Total Play < 10/15
Score > 75, Total Play > 10/15

Zimzat wrote:

I would like a way to skip, or at least stop, a song without affecting its score.

http://bugs.kde.org/show_bug.cgi?id=110298
But it is a bug, so enjoy it while it lasts:-p

Post edited by: mortiferus, at: 2005/08/10 21:39

This is actually a very complex issue, and I don\'t think there\'s an easy solution.

I\'ve been thinking a lot about how to do automatic ranking of songs given implicit information, like the user\'s listening patterns, and there\'s really no good way.

The problem is that the user\'s mood changes frequently, and the kind of songs they want to listen to change as well. The entire scoring system can be totally thrown off based on the user\'s mood -- I could\'ve scored all my songs when I was in a happy mood, then my girlfriend breaks up with me and my tastes are suddenly completely inverted!

An implicit scoring system would have to account for \"moods\" to be able to score songs properly. You could do this using data-mining and clustering algorithms to take a bunch of statistics, such as whether you were in front of your computer and how long you were there, as well as what songs you skipped and didn\'t skip (and even what songs you \"thumbs-up\"ed and \"thumbs-down\"ed), and then determine how your \"mood\" changes your taste.

For example, if I listened to a whole bunch of tracks in a row, and if I skipped a lot of moody depresing songs and didn\'t skip a lot of bouncy house music, it would cluster the house music in a new \"mood\" category. The mood doesn\'t have to be given a name; it\'s created automatically when the user\'s current taste in music doesn\'t fit in any existing mood (the user could name it themselves later though). If the songs that I skipped were already in their own category, then the \"mood\" can infer that those two moods are disjoint, and disassociate them.

The program would learn all of the different moods and their relationships (associations and disassociations) so that it could know that sometimes I want to listen to JUST house music, and sometimes JUST classical, and sometimes classical and house AND moody depressing music (rarely ). Groups and subgroups would be formed automatically.

The problem with this method, though, is that it takes a lot of user statistics to get accurate results. There are two solutions to this:
1) Use statistics from a whole community of users instead of having each new user retrain their own engine every time
2) Give the user more explicit input into the training process

The first solution is basically what audioscrobbler has been trying to do for a long time, and I think that people\'s tastes are so varied that it would be incredibly difficult to do accurately.

The second solution, however, could be quite powerful. One way you could implement it is by having the mood-engine use amarok\'s \"suggestions\" panel on the left to give potential songs that fit the user\'s current mood, and if the user wants to train the engine they can click thumbs-up and thumbs-down on each song. The engine should also be suggesting \"songs you probably don\'t want to hear\" along with \"songs you want to hear\", because it can make mistakes in that area too.

A SIMPLER MORE LOW TECH SOLUTION, of course, would be to add a slider to the program configuration that says how much of an effect skipping a song has on the score.

I use it when i am depressed (or not) and want to change a song without affecting the score. Just scroll the scrollbar (i mean, make the player play the last second or less) and then amaroK will play the next song, but affecting the score and playcount. It\'s what i use, maybe you aren\'t looking for somethin\' like this :unsure: ...

See you,

You guys are talking about rating and not scoring. Scoring is supposed to tell you what your long term favorites are, taking your moods into account. Rating is the manual ranking of songs based on a subjective assessment by the user, and you are all trying to use amaroK\'s scoring system as a rating system. if you are skipping a particular genre a lot because you are not in the mood, then you don\'t like those songs as much as the ones you are in the mood for.

I think a lot of people do this, and there has been some discussion in #amarok about fixing this, i.e. putting in a parallel rating system that the user can manually edit and stays static until the song rating is edited again.

The scoring system was designed for amaroK to tell you what you like over a long period of time, not for you telling amaroK what you like. It is unique and different and I think quite useful. I actually learned some things about my own listening habits by studying the results, and my new music purchases reflect those insights.