Hacker News Reader

Comment by tomc1985 | Hacker News Reader

tomc1985

Shuffle is great for playlists, but terrible for albums

I just wish shuffle was truly random, but it seems to favor tracks with the (edit- got this mixed up) lowest payout. I had the same track repeated several times to me in a few hours, in a playlist of several hundred tracks

View on HN

What you actually want is a true shuffle, where the entire set of songs is shuffled like a deck of cards, so you don’t get repeats. This is what iTunes does, to the best of my knowledge.

When you listen to 60 songs from a shuffled playlist of 300 songs, the likelihood of hearing the same song 3 or more times is approximately 29%. (Quick simulation: https://play.golang.org/p/2RNADWRQx4h)

This is related to the Birthday Paradox.

EDIT: Disregard this. I was thinking of shuffling as repeatedly choosing songs uniformly at random with replacement from the playlist, but commenters reminded me that shuffling shouldn't work that way. I found an open issue about this: https://community.spotify.com/t5/Desktop-Mac/Shuffle-should-...: "In a playlist of mine that is more than 6 hours long, shuffle often plays songs twice or three times within 30 minutes."

shuffle != random

Sure, hearing one track a few extra times I could accept.

But it happens a lot. And it seems to favor certain artists. I am well acquainted with the properties of randomness and what I've seen suggests some kind of weighting system.

https://engineering.atspotify.com/2014/02/28/how-to-shuffle-...

This is from 2014.

> it seems to favor tracks with the highest payout

Why would it do that? Wouldn't it optimize for the lowest per-stream rate?

You're right, I had my wires crossed.

Can confirm. I usually shuffle play my "liked songs" (which has nearly 1000 songs) and lately the same group of 40 or so songs play over and over. It's terrible.

That's exactly what I witnessed: a subset of the playlist receiving nearly all the attention

Reply on news.ycombinator.com