#231699 - 26/08/2004 18:08
The "random" debate continues
|
old hand
Registered: 20/03/2002
Posts: 729
Loc: Palo Alto, CA
|
Ahhhh the late-comers to mp3 players are now just starting to have the conversations we were having years ago. Is random really random? Enter the developers to say "You all are on crack. There are no patterns to shuffle. Random is random!"
Tunes, a Hard Drive and (Just Maybe) a Brain
_________________________
- trs
|
|
Top
|
|
|
|
|
#231701 - 26/08/2004 19:57
Re: The "random" debate continues
[Re: trs24]
|
carpal tunnel
Registered: 14/01/2002
Posts: 2858
Loc: Atlanta, GA
|
/me dusts off some old code . . .Too bad I couldn't read the article, though. 
_________________________
-Jeff
Rome did not create a great empire by having meetings; they did it by killing all those who opposed them.
|
|
Top
|
|
|
|
|
#231702 - 26/08/2004 20:21
Re: The "random" debate continues
[Re: trs24]
|
old hand
Registered: 28/12/2001
Posts: 868
Loc: Los Angeles
|
"The popularity of the listening mode led Apple's product design team to add Shuffle to the main menu on the fourth-generation iPod" Would the Karma developers please note this. 
_________________________
Ninti - MK IIa 60GB Smoke, 30GB, 10GB
|
|
Top
|
|
|
|
|
#231703 - 26/08/2004 22:15
Re: The "random" debate continues
[Re: trs24]
|
carpal tunnel
Registered: 12/11/2001
Posts: 7738
Loc: Toronto, CANADA
|
I like to keep a distinction between shuffle and random. Shuffle is not just random selection. Depending on implementation of course. With a rplain andom play you have the possibility of hearing the same exact file multiple times. It just keeps picking random track after random track. This is like shuffling a deck of cards and always putting the top card back into the deck at some random point (possible to hear the same track twice in a row). With "shuffle," you have a fixed number of tracks you want to randomly order, but each one plays once and only once. Like shuffling a deck of cards and always putting the top card into another pile. With the empeg, the argument is that the initial random ordering of the shuffle list isn't actually random. I've seen far too many coincidental pairings, but at least it's still what I think is a traditional shuffle mode. You won't get the same track again (unless you have two or more copies.... ) Bruno
|
|
Top
|
|
|
|
|
#231704 - 27/08/2004 23:06
Re: The "random" debate continues
[Re: hybrid8]
|
pooh-bah
Registered: 13/09/1999
Posts: 2401
Loc: Croatia
|
Quote:
With the empeg, the argument is that the initial random ordering of the shuffle list isn't actually random. I've seen far too many coincidental pairings, but at least it's still what I think is a traditional shuffle mode.
Depending on what you mean by 'actually random'. Arbitrary large shuffle without any perceived 'nonrandomness' (e.g. without two consecutive songs from the same album) is certainly not random. Appearance of any song in, say, position #2 in the list is equaly probable, including the one from the same album as the one in position #1. Forcing the songs from the same album (or any other specific pairs) 'apart' makes the ordering less random, not more.
That is not to say, of course, that what we actually want is mathematical randomness. We want appearance of randomness. Perhaps shuffle function should, after randomizing the running order, make an additional pass and further shuffle those song groups that would appear related (using some kind of proximity function). However, some occasional appearance of remnants of order will be very difficult to remove.
Did anybody measure how random the shuffle really is? (Say, make a short list of, say, 20 songs, two each from 10 albums. Calculate probabilty of songs from the same album appearing next to each other. Perform a large number of shuffles and count actual frequency of such pairs.) Didn't somebody do some kind of similar simulation a few years ago?
BTW, 'total disorder is impossible', as Ramsey theory finds out.
_________________________
Dragi "Bonzi" Raos
Q#5196
MkII #080000376, 18GB green
MkIIa #040103247, 60GB blue
|
|
Top
|
|
|
|
|
#231705 - 28/08/2004 06:50
Re: The "random" debate continues
[Re: bonzi]
|
pooh-bah
Registered: 02/06/2000
Posts: 1996
Loc: Gothenburg, Sweden
|
Randomness simulator was attached in this thread
_________________________
/Michael
|
|
Top
|
|
|
|
|
#231706 - 28/08/2004 15:51
Re: The "random" debate continues
[Re: bonzi]
|
carpal tunnel
Registered: 20/12/1999
Posts: 31597
Loc: Seattle, WA
|
Quote:
Did anybody measure how random the shuffle really is? (Say, make a short list of, say, 20 songs, two each from 10 albums. Calculate probabilty of songs from the same album appearing next to each other. Perform a large number of shuffles and count actual frequency of such pairs.) Didn't somebody do some kind of similar simulation a few years ago?
Yeah, here on the BBS, as mentioned, as well as in the FAQ here.
The discussion of how to create a "seems more random than actually random" shuffle is a good one. For example, deliberately spacing out the artists so they don't seem to play clusters.
I can think of some ways to create a shuffle mode where the artists and/or albums are spaced out evenly, but it has some problems. The main problem is that everything seems fine at the beginning of the shuffle, but as you get closer and closer to the end of the shuffle, the variation in track counts per artist starts to hurt you, and you can't help but get blocks of all the same artist. Also, artists who appear rarely on your player (for example, I have only two songs by Men at Work on my player) will appear near the beginning of every shuffle in that case.
|
|
Top
|
|
|
|
|
#231707 - 28/08/2004 20:11
Re: The "random" debate continues
[Re: bonzi]
|
pooh-bah
Registered: 19/09/2002
Posts: 2494
Loc: East Coast, USA
|
Hm... my unscientific done-while-driving-75-on-New-Jersey-highways analysis of the empeg random leads me to the conclusion: I like it. It's not "random", it's more of "a good mix". It seems to try and pick from different playlists the majority of the time, picking from the same occasionally. I'd love to know how the assumed logic in the previous sentence applies to randomizing a playlist with nested playlists. I could explain it with a drawing better than in words, so I'll just pas (also because my assumption of the randomization is probably totally wrong). Either way, empeg random always gives me "a good mix."
_________________________
- FireFox31 110gig MKIIa (30+80), Eutronix lights, 32 meg stacked RAM, Filener orange gel lens, Greenlights Lit Buttons green set
|
|
Top
|
|
|
|
|
#231708 - 30/08/2004 15:04
Re: The "random" debate continues
[Re: tfabris]
|
carpal tunnel
Registered: 14/01/2002
Posts: 2858
Loc: Atlanta, GA
|
Quote:
The discussion of how to create a "seems more random than actually random" shuffle is a good one.
What worked out well for me for simple mp3 player I wrote once (in the desktop world) was to create a random running list, then loop through it looking for "dupicates". When I found one I'd move the offending song a fixed number of spaces away from the first one. Yes, this did end up clustering dense artists at the end, but with a mix of thousands of songs I rarely get there anyway. What this doesn't do is force artists you have only a few songs from to always show up near the beginning (in an effort to evenly space things). For something like the Karma where you pick the "entertain me for X minutes", you could improve this idea by creating a full running list, REMOVING the duplicates, and then choping off the playlist at the perscribed time. Then you wouldn't get the clustering at the end.
_________________________
-Jeff
Rome did not create a great empire by having meetings; they did it by killing all those who opposed them.
|
|
Top
|
|
|
|
|
|
Previous Topic
Index