Multivariate Weighted Shuffle
After Apple had been propagating iTunes for a few versions, they (at length) released a version that allows its DJ function a binary option that instructs iTunes to “play higher rated songs more often.” Break out the champagne, everyone! Woohoo! The sad thing is that this still amounted to a large step forward in enhancing apple’s random Shuffle setting.
(“Random,” that is, unless you’re one of the believer’s in steve Jobs’s supernatural ability to dictate iTunes to play particular sorts of song more or less often. This used to crop up every now and then, when people who don’t distinguish pareidolia from non-random machine behaviour. Much as I am surprised by the occasions when the entirely separate shuffle routines of my home computer and my work computer turn up the same track on the same day — which really is quite striking, since each has a very, very large database of tracks to choose from — I do not suspect that Apple’s developers have figured out a way to associate my two computers, transmit data about tracks it would surprise me to hear from both, and keep those coincidences subtle and rare enough that they remain unpredictable.)
Still, one switch that assigns one (undefined) weight to one variable is very far from a twenty-first century solution to the “weighted shufflke” problem. Since there’s already an “Advanced” menu for the app, and a small “Settings” button in the iTunes DJ view, why couldn’t one of these options lead to a “weighted Shuffle” option that’s more like a Smart Playlist or a Spotlight search? Why, in other words, can’t I set my iTunes DJ to select songs with a one-to-five weighting of multivariate criteria? If a vanilla random track is weighted at “one,” I’d like
Songs from the “Recently Added” list to have a weighted prominence of 3
Songs by women artists* to have a weighted prominence of 2 (or 3)
Songs that iTunes’s “Genius” function associates with the current track to have a weight of 4
Songs from 1967 to 1985 to have a weight of 3
Songs from 2000 to 2008 to have a weight of 3
Songs from the last two years to have a weight of 4
Songs from 1955 to 1966 to have a weight of 2
Songs that have never been played to have a weight of 5
Songs that have been played only once have a weight of 4
Songs that have been played twice have a weight of 2
Songs rated 5 stars have a weight of 5 (and so on)
Songs by Scottish artists to have a weight of 2
So that a five-star track that I haven’t listened to before (let’s assume I already know that I like it from Pippa playing it for me), just released this year, that I’ve just downloaded from eMusic, by a Scottish woman artist would be very highly likely to turn up on my iTunes DJ in the near future — whereas a track that I’ve heard four or five times, that I don’t ever like that much (say, two stars), from the Dark Interval between 1986 and 2000, that I acquired a couple of years ago, by a male American artist, would be proportionately unlikely to come up any time soon.
The database operations to support that — and even more complex weighting — must be much less onerous than programming most of the eye candy visualizers that Apple distributes, and it would change the iTunes DJ from a mostly-random function to something much more like a personal genius-DJ. Since I listen to the iTunes DJ playlist almost exclusively, I’d like this upgrade a whole lot.**
* This would take some finagling with the current iTunes database, but I could use the “Genre” category (which I otherwise ignore entirely) to make a category for women artists, or could set up a playlist, songs that appear on which would be double-weighted. Same with Scottish artists.
** I realise I might be able to fudge an iTunes DJ list that came somewhat close to this function by rigging an elaborate network of interconnected playlists — I already do this to enhance the Shuffle function on my iPhone — but honestly, it would be vastly easier for Apple to make the database categories more useful than for each user to rig a work-around.