Comment Non-random shuffle play (Score 1) 718
This thread raises one of my ongoing pet peeves with the IPod and similar devices: the random number generators used for shuffle play tend to be very low quality. They tend to exhibit a substantial amount of serial correlation.
This is evident when you shuffle a group of approximately 2^n songs. There is a distinct tendency for the shuffle list to revisit specific parts of the song list at regular intervals. Thus, if you hear a song by a certain group or artist, there is a high probability that you will hear another song by the same artist/group shortly thereafter.
There are simple ways to reduce serial "non-random" behavior of pseudo-random generators in code (take a look at Numerical Recipes for a few good algorithms), but nobody seems to be aware of the problem: all pseudo-random generators are not created equal!
This is evident when you shuffle a group of approximately 2^n songs. There is a distinct tendency for the shuffle list to revisit specific parts of the song list at regular intervals. Thus, if you hear a song by a certain group or artist, there is a high probability that you will hear another song by the same artist/group shortly thereafter.
There are simple ways to reduce serial "non-random" behavior of pseudo-random generators in code (take a look at Numerical Recipes for a few good algorithms), but nobody seems to be aware of the problem: all pseudo-random generators are not created equal!
