Predicting songs on shuffle

S

slackfarm

New member
Joined
Aug 25, 2019
Messages
5
Assuming you have an iPod4 with 500 songs, and you put them all on shuffle play, how many songs would you have to observe before you could mathematically determine the shuffle/PRNG algorithms and be able to predict the next song with 100% accuracy?

I don't know Apple's specifics, and they won't tell me, but I assume the simplest rule set for shuffling would be something like: [in a list of T songs, let N=1. Choose a random integer X between (and including) T and N. Then swap the list items at positions N and X.]

The most common PRNG (or pseudo random number generator), according to Google, seems to be the Linear Conguential Generator.
 
1) If it's as deficient as you are expecting, it has a period. If you know the period, you're done.

2) Are the songs actually as ORDERED as you seem to think they are?

3) There's a whole lot more out there than PRNG and LCG. Your likelihood of picking the right kind of generator is negligible unless it's particularly simplistic.

4) There is no such thing as "predict" with "100% accuracy". That's just a fixed outcome; no prediction about it.

5) What's the point of "shuffle" if you know what's next?
 
tkhunny said:
1) If it's as deficient as you are expecting, it has a period. If you know the period, you're done.

2) Are the songs actually as ORDERED as you seem to think they are?

3) There's a whole lot more out there than PRNG and LCG. Your likelihood of picking the right kind of generator is negligible unless it's particularly simplistic.

4) There is no such thing as "predict" with "100% accuracy". That's just a fixed outcome; no prediction about it.

5) What's the point of "shuffle" if you know what's next?
Click to expand...

I did a study where test subjects listened to shuffled iPods with 25, 50, 100 and 500 songs. I recorded how long it took subjects to accurately predict the next song, and the degree of accuracy. This was intuitive human prediction and I want to compare its efficacy versus "cracking the algorithm" with math.
 
I don't get it. See particularly point #4, above.
 
If you listen to 20 songs on shuffle for about 10 hours, the average person will have identified a pattern, and will be able to yell you the next song to be played in that pattern. I want to know how you would go about identifying that same pattern with math. Or how one might code an app to identify the pattern. Since Apple won't tell me how the iPod shuffles, I'm left with no other option.
 
In your study i really hope you also records the songs, the appearance of each song compare to others should be the same right? Or was it different? The probability of 100% or accuracy of 100% is not achievable with random set, its random, unless the random thing is not that random.
 
It's not random. Computers are not capable of generating truly random numbers. Their pseudo random number generation is always some kind of algorithm.
 
slackfarm said:
It's not random. Computers are not capable of generating truly random numbers. Their pseudo random number generation is always some kind of algorithm.
Click to expand...
I asked other questions too, please answer them ;)
 
You must log in or register to reply here.
Top