this blog is girtby.net

Posted
03 November 2005

Categories
Nerd Factor X

Tags

7 Comments

iTunes Library Preening

Preening one's iTunes library is one of those continually rewarding pastimes. Adding cover art, fixing up meta-data, adding playlists, it's all endless fun. iTunes has a particularly entertaining feature where each track in your library can be given a rating, from one to five stars. This is a great way of instantly locating the best-of-the-best tracks in your collection. If you're like me you preen your ratings constantly.

The trouble I've always had though is deciding on the criteria for a given rating. Here are the criteria I have been loosely applying up till now. First the easy ones, then the harder ones:

  • 5 Stars Desert-island track. Listen anytime, anywhere. Ridiculous sentimental attachment.
  • 1 Star A bit crap. Sometimes totally crap, but I just can't bear to part with it because it completes an otherwise worthy album. Maybe it's one of those "we're just tooling around with the microphones on but we'll put it on the album for a laugh" type of tracks. On shuffle-play I don't want to hear these, ever.
  • 2 Stars A default rating. Nothing standout about this track. Could stand to lose it from the collection, but don't actually dislike it.
  • 4 Stars Something about this prevents it accompanying me to the desert island. Maybe I only like it in certain circumstances, or maybe I used to love it but am now going a bit bored of it.
  • 3 Stars Somewhere in-between 2 and 4. The less said about this category the better, OK?

After struggling with these definitions for a while, I decided that they weren't really working. The 2, 3, and 4 star ratings were the hard ones, and this is where the bulk of my ratings go. This led me to thinking about the distribution of the tracks with each rating. Which led to the idea that maybe the ratings should reflect the distribution.

So instead of defining 5 stars against some arbitrary external criteria, it seems to be just as useful (if not more so) to define these in relative terms. So instead of "desert island track", I now look at 5 star-rated tracks as those in the top 5% of my library. 4 stars are in the top 15%, 3 stars in the top 50%, and 2 stars in the top 90%. 1 Star reserved for the bottom 10%.

[Don't ask me why 5 stars is the top 5% and 1 star is the bottom 10%, they are just the figures that popped out of my brain, perhaps some subconsciously-mangled Sturgeon's Law?]

But I needed some help to assess how far away from these definitions in my current ratings assignments. So I wrote a script to analyse my iTunes library to determine how far away I was from my target ratings distribution. Here is the first result:

4339 tracks, 613 (14%) rated
Cumulative % of rated
Number % of rated Actual Target Shortfall
Tracks rated 5 stars: 6 1 1 5 4
Tracks rated 4 stars: 136 22 23 15 -8
Tracks rated 3 stars: 277 45 68 50 -18
Tracks rated 2 stars: 160 26 94 90 -4
Tracks rated 1 stars: 34 6 100

As you can see, I either need to realign my ratings definitions, change some of the existing assignments, or overcompensate when rating new tracks. Or all of these. Anyway, I don't want to be seen to be obsessing over this too much (a bit late now, he realises!) but the idea is to see that I've been, for example, a bit stingy with the 5-star ratings, and a bit generous with the 3-star ratings.

All of this obsessive behaviour can be yours, thanks to the latest girtby offering: iTunes Library Stats. It's an XSLT script which will produce the above output for your own iTunes Library. Works on Windows and MacOS. Enjoy!

7 Comments

Posted by
Chris
2005-11-03 21:45:52 -0600
#

What about unrated tracks? Or do you obsessively rate everything?


Posted by
alastair
2005-11-03 21:45:52 -0600
#

Unrated tracks are ignored, hence "% of rated" and "cumulative % of rated".


Posted by
Eddie
2005-11-03 21:45:52 -0600
#

It's interesting to see someone think about iTunes track rating the way I do. I haven't found a good method yet (mine matches what you orginally had plus or minus).

The problem I have is that I have tracks that I would consider 5 star, but not for the particular circumstances.. ie- the gym, vs. sleeping. I see you give those a 4. But then I get (as you have) very few 5 stars that I would listen to any time, anywhere.

That's pretty much why I stopped obsessively rating them- I now have complex smart playlists. So complex, that I would welcome the addition of regular expressions to further granulate :)


Posted by
james
2005-11-03 21:45:52 -0600
#

Honestly, it doesn't make any sense to me that one star be the rating for dislike. To me, if I don't like something enough to give it a negative rating it is most likely not in music library to begin with. Therefor, it makes more sense to me to use the 5 star rating system as a way of indicating increasing level of satisfaction with the song. One star being this song is a cut above the great majority which I am indiffernt to and there for get no ranking at all. Two through four stars are simply increasingly preferred songs, with five stars being my most prefered songs.

My concern/confussion is in how iTunes actually makes this distinction in how it calculates shuffle play. Does iTunes play songs ranked with one star more often than songs with no stars?


Posted by
alastair
2005-11-03 21:45:52 -0600
#

Eddie, the underrated "grouping" field is great for this kind of task. The 1-5 star rating doesn't have a lot of resolution for saying "well I love this track except that it really is too rockin' to play when I'm brushing my teeth in the morning, except if I've just been for a run beforehand, and ..." So IMHO you're better off just using the rating to say how much you like the track. Categorising it to certain situations is the job of playlists (smart or dumb).


Posted by
alastair
2005-11-03 21:45:52 -0600
#

James:

Well perhaps I should point out that I have a 60GB iPod so I have plenty of space for the one star tracks...

Let me use an example which you might be familiar with: Radiohead's "OK Computer" album. Fantastic album. For whatever reason I resisted its charms, but now I'm totally into it. Now a lot of people (not including me BTW) dislike the track called "Fitter Happier". You might know the one, it has a robot voice reciting self-improvement goals. Some people don't like it and skip the track reflexively when it comes on.

So this hypothetical person (again not me, I quite like the track although it's by no means the best) might rate this as one star to indicate that they don't really like it. They could instead just delete the song entirely from the library, but this leaves something of a gap in the album. I don't know about you, but I have a bit of an all-or-nothing approach, so I would prefer to keep the track but mark it as not a desirable candidate for shuffle play. If I wanted to skip the track during normal (sequential) play, I would probably create a smart playlist that excluded the one star tracks, or even a regular playlist with that track excluded.

iTunes calculation of shuffle play is most certainly a mystery. Particularly in version 5.0+ where they introduced the "smart shuffle" feature. It is a mystery that will remain unsolved (by me).


Posted by
james
2005-11-03 21:45:52 -0600
#

alastair, Good point and well taken. I too am reluctant to break apart album, although couldn't the check marks have the same effect? I believe you can you not uncheck songs that you want not to be generally played while still keeping them in the library. There is an option when making smart playlists to only include checked songs. I think maybe I have a mental deficiency when it comes to star rating systems. If I knew how the iTunes shuffle worked I could be more comfortable with a ratings system any which way.