- Joined
- Sep 11, 2019
- Messages
- 7,592
- Credits
- 7,310
![☆ ITZY - LOCO [Ver. C] ☆](/forums/data/dbtechShop/itemIcons/2/2002.jpg?1632758575)
Just before we begin, I would like to mention when I started the experiment I did not consider the LOONA boycott. Because of wanting to keep the variables consistent all week, I had to leave LOONA's songs on the playlists. But I promise, from now on, no LOONA. I didn't listen to them that much over the week anyway
Have you ever had a playlist, big or small, and you could swear on your life every time you shuffle said playlist, the same songs always play?
This is something I know I've experienced. And I did actually find a video that analysed the Spotify shuffle algorithm and experimented with the shuffle feature using a smaller playlist of like under 50 songs iirc. The thing is, my main playlist I listen to every morning is a LOT bigger than 50 songs... like over 10x bigger. So I was curious how this would look on a playlist of that size.
Methodology
In order to my results to not be effected by songs entering and leaving my playlist, as this may have been a variable, I decided to keep the same 605 songs on my playlist all week. This would ensure the same 605 songs all had the exact same change to be shuffled each day.
Every morning when I woke up and shuffled my playlist, I took a screenshot of my queue for the first 30 songs in that queue. I chose 30 songs because the length of time I listened to my playlist varied each day so 30 felt like a nice number.
At the end of the week, I became sick... joking. Well... not the sick part. But I did analyse the data I collected. And now I shall reveal the results.
Another thing to note is my Spotify is weird at the moment, and so if I leave my queue unattended for too long it resets. So if this happened but I wanted to keep listening, I reshuffled the playlist and got those 30 songs ON TOP OF the 30 already shuffled.
What I calculated
There were a selection of things I wanted to find out from this:
1. How many songs on average did I listen to on the playlist per day
2. How frequently my most common artists on the playlist play over the week
3. Which songs appeared more than once
Statistics before the findings
Over the 7 days, I shuffled my playlist a total of 10 times, shuffling twice on Monday, Wednesday and Thursday
10 shuffles with 30 songs each means I was working with a sample of 300 songs
300 is just under 50% of the total playlist length, falling at 49.6% to one d.p
Theoretically, this means I could have listened to up to 300 songs. But I haven't done statistics since Year 11, which was 1 and a half years ago. So sadly I cannot give you the probability of a specific song appearing twice
I made a list of all my artists who have at least 8 songs on that playlist, found the percentage of how many songs they occupy in the full playlist, and used that percentage to make an estimate of how many times we'd expect to see that artist within the final 300 songs. (Note, I included any sub-unit song that had the group name in the title, eg WJSN Chocome. This was out of laziness)
BRAVE GIRLS, SONAMOO - 8 songs, 1.32%, expected 4 times
WEEEKLY - 9 songs, 1.49%, expected 5 times
EVERGLOW, LOONA, Purple Kiss - 10 songs, 1.65%, expected 5 times
CLC, Oh My Girl, Stray Kids, Weki Meki - 11 songs, 1.82%, expected 6 times
Dreamcatcher - 12 songs, 1.98%, expected 6 times
ATEEZ - 17 songs, 2.81%, expected 9 times
MAMAMOO - 18 songs, 2.98%, expected 9 times
Red Velvet - 20 songs, 3.31%, expected 10 times
AOA - 22 songs, 3.64%, expected 11 times
TWICE - 32 songs, 5.29%, expected 16 times
ITZY - 34 songs, 5.62%, expected 17 times
WJSN - 35 songs, 5.79%, expected 18 times (everyone gasping coz ITZY isn't first)
WHAT I FOUND
First of all, the boring statistic. On average, I listened to 20.7 songs (so basically 21 songs) each day before I switched playlist (or stopped counting the queue). Meaning I listened to approximately 70% of the songs.
Okay, now for the fun.
Of the 300 recorded songs, 52 of them appeared more than once. Overal 244 unique songs appeared, which makes up 81.3% of the 300 sample, or 40.3% of the total 605 songs. So within a week, I didn't have the possibility of even listening to 50% of my whole playlist.
48 of these duplicate songs appeared twice, and 4 of them appeared 3 times total.
The 3 songs that appeared 3 times in the queues were:
Mi Gente by J Balvin, which is a latina song so yeah one of the songs that appeared the most wasn't even kpop. Surprise
GAMBLER - Monsta X
VIVACE - LIGHTSUM
and Ponzona by Purple Kiss
Furthermore, duplicates from the list of expected artists go as followed
1 duplicate: AOA, LOONA, Oh My Girl, Weeekly
2 duplicates: MAMAMOO, Purple Kiss
3 duplicates: ATEEZ, Red Velvet, WJSN
4 duplicates: ITZY, TWICE
As for the number of times the predicted artists appeared, here is a list going from they appeared less to they appeared more:
-3 TIMES: AOA, CLC, Stray Kids
-2 TIMES: Brave Girls, Oh My Girl, SONAMOO, Weeekly, Weki Meki
-1 TIME: DREAMCATCHER, EVERGLOW, MAMAMOO, TWICE
EXACT TIMES PREDICTED: WJSN
+1 TIME: ATEEZ, LOONA
+2 TIMES: Red Velvet
+3 TIMES: ITZY
+5 TIMES: Purple Kiss
Overall, out of the artists I was monitoring, ITZY appeared the most times, having 20 songs appear in the queue
Brave Girls and SONAMOO appeared the least, only appearing twice
WHAT I HAVE LEARNT
From this experiment, I have learnt that some songs do seem to appear more often than others, though it also seems like the songs that do appear often are quite random, as none of the 4 that appeared were ones I was like losing my mind over because I wanted to listen to. So it wasn't the algorithm trying to push them based on my listening habits.
What I found quite interesting is the appearance predictions were actually quite similar to the final numbers. Though of course, listing them based on if they were a plus or minus isn't the best way of showing them, as for example SKZ appeared 3 less times than expected, but that was half the number of times, whereas AOA also appeared 3 less times, yet that was still 72.7% of the expected times.
I also learnt that no matter if WJSN thinks they're on top, ITZY will always find a way.
Thanks for reading! I hope you found this interesting coz I got so bored of this experiment by the end of the week and I do not miss it. So please don't let my suffering be useless
Have you ever had a playlist, big or small, and you could swear on your life every time you shuffle said playlist, the same songs always play?
This is something I know I've experienced. And I did actually find a video that analysed the Spotify shuffle algorithm and experimented with the shuffle feature using a smaller playlist of like under 50 songs iirc. The thing is, my main playlist I listen to every morning is a LOT bigger than 50 songs... like over 10x bigger. So I was curious how this would look on a playlist of that size.
Methodology
In order to my results to not be effected by songs entering and leaving my playlist, as this may have been a variable, I decided to keep the same 605 songs on my playlist all week. This would ensure the same 605 songs all had the exact same change to be shuffled each day.
Every morning when I woke up and shuffled my playlist, I took a screenshot of my queue for the first 30 songs in that queue. I chose 30 songs because the length of time I listened to my playlist varied each day so 30 felt like a nice number.
At the end of the week, I became sick... joking. Well... not the sick part. But I did analyse the data I collected. And now I shall reveal the results.
Another thing to note is my Spotify is weird at the moment, and so if I leave my queue unattended for too long it resets. So if this happened but I wanted to keep listening, I reshuffled the playlist and got those 30 songs ON TOP OF the 30 already shuffled.
What I calculated
There were a selection of things I wanted to find out from this:
1. How many songs on average did I listen to on the playlist per day
2. How frequently my most common artists on the playlist play over the week
3. Which songs appeared more than once
Statistics before the findings
Over the 7 days, I shuffled my playlist a total of 10 times, shuffling twice on Monday, Wednesday and Thursday
10 shuffles with 30 songs each means I was working with a sample of 300 songs
300 is just under 50% of the total playlist length, falling at 49.6% to one d.p
Theoretically, this means I could have listened to up to 300 songs. But I haven't done statistics since Year 11, which was 1 and a half years ago. So sadly I cannot give you the probability of a specific song appearing twice
I made a list of all my artists who have at least 8 songs on that playlist, found the percentage of how many songs they occupy in the full playlist, and used that percentage to make an estimate of how many times we'd expect to see that artist within the final 300 songs. (Note, I included any sub-unit song that had the group name in the title, eg WJSN Chocome. This was out of laziness)
BRAVE GIRLS, SONAMOO - 8 songs, 1.32%, expected 4 times
WEEEKLY - 9 songs, 1.49%, expected 5 times
EVERGLOW, LOONA, Purple Kiss - 10 songs, 1.65%, expected 5 times
CLC, Oh My Girl, Stray Kids, Weki Meki - 11 songs, 1.82%, expected 6 times
Dreamcatcher - 12 songs, 1.98%, expected 6 times
ATEEZ - 17 songs, 2.81%, expected 9 times
MAMAMOO - 18 songs, 2.98%, expected 9 times
Red Velvet - 20 songs, 3.31%, expected 10 times
AOA - 22 songs, 3.64%, expected 11 times
TWICE - 32 songs, 5.29%, expected 16 times
ITZY - 34 songs, 5.62%, expected 17 times
WJSN - 35 songs, 5.79%, expected 18 times (everyone gasping coz ITZY isn't first)
WHAT I FOUND
First of all, the boring statistic. On average, I listened to 20.7 songs (so basically 21 songs) each day before I switched playlist (or stopped counting the queue). Meaning I listened to approximately 70% of the songs.
Okay, now for the fun.
Of the 300 recorded songs, 52 of them appeared more than once. Overal 244 unique songs appeared, which makes up 81.3% of the 300 sample, or 40.3% of the total 605 songs. So within a week, I didn't have the possibility of even listening to 50% of my whole playlist.
48 of these duplicate songs appeared twice, and 4 of them appeared 3 times total.
The 3 songs that appeared 3 times in the queues were:
Mi Gente by J Balvin, which is a latina song so yeah one of the songs that appeared the most wasn't even kpop. Surprise
GAMBLER - Monsta X
VIVACE - LIGHTSUM
and Ponzona by Purple Kiss
Furthermore, duplicates from the list of expected artists go as followed
1 duplicate: AOA, LOONA, Oh My Girl, Weeekly
2 duplicates: MAMAMOO, Purple Kiss
3 duplicates: ATEEZ, Red Velvet, WJSN
4 duplicates: ITZY, TWICE
As for the number of times the predicted artists appeared, here is a list going from they appeared less to they appeared more:
-3 TIMES: AOA, CLC, Stray Kids
-2 TIMES: Brave Girls, Oh My Girl, SONAMOO, Weeekly, Weki Meki
-1 TIME: DREAMCATCHER, EVERGLOW, MAMAMOO, TWICE
EXACT TIMES PREDICTED: WJSN
+1 TIME: ATEEZ, LOONA
+2 TIMES: Red Velvet
+3 TIMES: ITZY
+5 TIMES: Purple Kiss
Overall, out of the artists I was monitoring, ITZY appeared the most times, having 20 songs appear in the queue
Brave Girls and SONAMOO appeared the least, only appearing twice
WHAT I HAVE LEARNT
From this experiment, I have learnt that some songs do seem to appear more often than others, though it also seems like the songs that do appear often are quite random, as none of the 4 that appeared were ones I was like losing my mind over because I wanted to listen to. So it wasn't the algorithm trying to push them based on my listening habits.
What I found quite interesting is the appearance predictions were actually quite similar to the final numbers. Though of course, listing them based on if they were a plus or minus isn't the best way of showing them, as for example SKZ appeared 3 less times than expected, but that was half the number of times, whereas AOA also appeared 3 less times, yet that was still 72.7% of the expected times.
I also learnt that no matter if WJSN thinks they're on top, ITZY will always find a way.
Thanks for reading! I hope you found this interesting coz I got so bored of this experiment by the end of the week and I do not miss it. So please don't let my suffering be useless
