Comments on: Fun with statistical clustering http://blog.xcski.com/2005/10/12/fun-with-statistical-clustering Everything I used to bore people on newsgroups and mailing lists with, now in one inconvenient place. Fri, 05 Dec 2008 10:45:29 +0000 http://wordpress.org/?v=2.6.5 By: Tina Marie http://blog.xcski.com/2005/10/12/fun-with-statistical-clustering#comment-1466 Tina Marie Wed, 12 Oct 2005 22:38:14 +0000 http://xcski.com/blogs/pt/2005/10/12/fun-with-statistical-clustering#comment-1466 I don't think it's you. My iPod has a definate preference for some songs/artists/albums. I ended up defining all my playlists as "sort by least played", and it's helped a bit, but shuffle still isn't random for me. I hear the new iTunes makes shuffle more configurable. I don’t think it’s you. My iPod has a definate preference for some songs/artists/albums. I ended up defining all my playlists as “sort by least played”, and it’s helped a bit, but shuffle still isn’t random for me.

I hear the new iTunes makes shuffle more configurable.

]]> By: Laura http://blog.xcski.com/2005/10/12/fun-with-statistical-clustering#comment-1465 Laura Wed, 12 Oct 2005 22:17:44 +0000 http://xcski.com/blogs/pt/2005/10/12/fun-with-statistical-clustering#comment-1465 I think the iPods just prefer to play certain artists or songs. Mine almost always goes through my entire Showtunes playlist before it plays anything from Rent. I think the iPods just prefer to play certain artists or songs. Mine almost always goes through my entire Showtunes playlist before it plays anything from Rent.

]]> By: Mentis Fugit http://blog.xcski.com/2005/10/12/fun-with-statistical-clustering#comment-1460 Mentis Fugit Wed, 12 Oct 2005 20:17:23 +0000 http://xcski.com/blogs/pt/2005/10/12/fun-with-statistical-clustering#comment-1460 Just looking at one of those examples, I figure the odds of hearing two Shonen Knife songs in one day, given you've heard one (got to have one before you can discern a "pattern" after all) thus: The chance a song is <i>not</i> by Shonen Knife is going to be (10,000 - 46) / 10,000 = 0.9954. The chance that all of the other 99 songs are <i>not</i> by Shonen Knife is going to be 0.9954^124 = 0.63. If I'm thinking straight, the odds you'll hear three are 0.37 * (1 - 0.9954^98) = 0.14. If Shonen Knife's tally of 46 tracks were representative of other artists on your ying-tong-iddle-iPod, roughly similar odds would apply to every other artist. In other words, once you'd heard one song by X, the chances of hearing two more are about one in seven. With presumably over 200 different artists represented... ...no, I can't remember how to work it out, but the odds are better than you think. Just looking at one of those examples, I figure the odds of hearing two Shonen Knife songs in one day, given you’ve heard one (got to have one before you can discern a “pattern” after all) thus:

The chance a song is not by Shonen Knife is going to be (10,000 - 46) / 10,000 = 0.9954.

The chance that all of the other 99 songs are not by Shonen Knife is going to be 0.9954^124 = 0.63.

If I’m thinking straight, the odds you’ll hear three are 0.37 * (1 - 0.9954^98) = 0.14.

If Shonen Knife’s tally of 46 tracks were representative of other artists on your ying-tong-iddle-iPod, roughly similar odds would apply to every other artist. In other words, once you’d heard one song by X, the chances of hearing two more are about one in seven. With presumably over 200 different artists represented…

…no, I can’t remember how to work it out, but the odds are better than you think.

]]> By: rone http://blog.xcski.com/2005/10/12/fun-with-statistical-clustering#comment-1459 rone Wed, 12 Oct 2005 20:05:33 +0000 http://xcski.com/blogs/pt/2005/10/12/fun-with-statistical-clustering#comment-1459 Yep. I combat this by listening to a smart playlist that only contains songs with a play count under 2. Yep. I combat this by listening to a smart playlist that only contains songs with a play count under 2.

]]>