CVGM.net - Chiptune, Retro & Video Game Music

Hi,

I have been fiddling with some new queue changes lately, and a few people have started noticing it, so before I forget to talk about it again, i'll post some info about it here.

A while ago on Nectarine, sqrmax came up with an algo that computes locktimes on songs based on various criteria, including number of votes, overall rating and length of song. The function uses an S curve to compute the best values based on the information that it has. The purpose of this is to take the good songs, and play them more often, and take the songs that are repeatedly rated badly, and push them back out so they take longer to appear in the queue again.

For anyone interested in the script, or has time and patience to mess around with numbers, can grab the (modified) file from here:
http://www.cvgm.net/static/media/misc/song_locktime_cvgm.py

The output:

Song name Rating Votes Length Magic number Lock Time
No ratings, length bias A None 0 100 0.238017398306 D:38 H:14 M:10 S:23
No ratings, length bias B None 0 200 0.238017398306 D:38 H:14 M:10 S:23
No ratings, length bias C None 0 350 0.241425309996 D:38 H:19 M:6 S:3
No ratings, length bias D None 0 500 0.246155586717 D:39 H:1 M:56 S:27
No ratings, length bias E None 0 1200 0.26868882153 D:40 H:10 M:31 S:26
No ratings, length bias F None 0 1500 0.278059390198 D:41 H:0 M:4 S:25
No ratings, length bias G None 0 1800 0.286919163612 D:41 H:12 M:53 S:6
Benefit of the doubt 1A None 0 312 0.24024153336 D:38 H:17 M:23 S:21
Benefit of the doubt 1B 4.5 1 312 0.222262562865 D:37 H:15 M:23 S:29
Benefit of the doubt 1C 4.5 2 312 0.204389785065 D:36 H:13 M:32 S:51
Benefit of the doubt 1D 4.5 3 312 0.186628814082 D:35 H:11 M:51 S:54
Benefit of the doubt 1E 4.5 4 312 0.168985030714 D:34 H:10 M:21 S:8
Benefit of the doubt 1F 4.5 5 312 0.151463576077 D:33 H:9 M:0 S:58
Benefit of the doubt 1G 4.5 6 312 0.134069346351 D:32 H:7 M:51 S:51
Benefit of the doubt 1H 4.5 7 312 0.116806988636 D:31 H:6 M:54 S:10
Benefit of the doubt 1I 4.5 8 312 0.117490107427 D:31 H:7 M:53 S:26
Benefit of the doubt 1J 4.5 9 312 0.117961686888 D:31 H:8 M:34 S:21
Benefit of the doubt 1K 4.5 10 312 0.11830057242 D:31 H:9 M:3 S:45
Benefit of the doubt 1L 4.5 15 312 0.119108237789 D:31 H:10 M:13 S:49
Benefit of the doubt 1M 4.5 20 312 0.119392543892 D:31 H:10 M:38 S:29
Benefit of the doubt 1N 4.5 35 312 0.119639380795 D:31 H:10 M:59 S:54
Benefit of the doubt 1O 4.5 50 312 0.119700509875 D:31 H:11 M:5 S:12
Benefit of the doubt 1P 4.5 75 312 0.119733153664 D:31 H:11 M:8 S:2
Benefit of the doubt 1Q 4.5 100 312 0.119744581447 D:31 H:11 M:9 S:2
Benefit of the doubt 2A None 0 312 0.24024153336 D:38 H:17 M:23 S:21
Benefit of the doubt 2B 1.5 1 312 0.331517485351 D:44 H:5 M:22 S:27
Benefit of the doubt 2C 1.5 2 312 0.424488164376 D:49 H:19 M:48 S:35
Benefit of the doubt 2D 1.5 3 312 0.518253589059 D:55 H:11 M:23 S:40
Benefit of the doubt 2E 1.5 4 312 0.611866589525 D:61 H:2 M:45 S:32
Benefit of the doubt 2F 1.5 5 312 0.704389137881 D:66 H:16 M:32 S:48
Benefit of the doubt 2G 1.5 6 312 0.794947341552 D:72 H:3 M:29 S:37
Benefit of the doubt 2H 1.5 7 312 0.882777310031 D:77 H:10 M:29 S:45
Benefit of the doubt 2I 1.5 8 312 0.883458191487 D:77 H:11 M:28 S:49
Benefit of the doubt 2J 1.5 9 312 0.883921617176 D:77 H:12 M:9 S:2
Benefit of the doubt 2K 1.5 10 312 0.884251369721 D:77 H:12 M:37 S:38
Benefit of the doubt 2L 1.5 15 312 0.885026443869 D:77 H:13 M:44 S:53
Benefit of the doubt 2M 1.5 20 312 0.885295709926 D:77 H:14 M:8 S:15
Benefit of the doubt 2N 1.5 35 312 0.88552800601 D:77 H:14 M:28 S:24
Benefit of the doubt 2O 1.5 50 312 0.8855853226 D:77 H:14 M:33 S:22
Benefit of the doubt 2P 1.5 75 312 0.88561589622 D:77 H:14 M:36 S:2
Benefit of the doubt 2Q 1.5 100 312 0.88562659366 D:77 H:14 M:36 S:57
Benefit of the doubt 3A None 0 312 0.24024153336 D:38 H:17 M:23 S:21
Benefit of the doubt 3B 3.0 1 312 0.276494475342 D:40 H:21 M:48 S:39
Benefit of the doubt 3C 3.0 2 312 0.313099343686 D:43 H:2 M:44 S:29
Benefit of the doubt 3D 3.0 3 312 0.350003731381 D:45 H:8 M:6 S:19
Benefit of the doubt 3E 3.0 4 312 0.387152485929 D:47 H:13 M:49 S:20
Benefit of the doubt 3F 3.0 5 312 0.424488164376 D:49 H:19 M:48 S:35
Benefit of the doubt 3G 3.0 6 312 0.461951540827 D:52 H:1 M:58 S:54
Benefit of the doubt 3H 3.0 7 312 0.499482155855 D:54 H:8 M:15 S:4
Benefit of the doubt 3I 3.0 8 312 0.501181298292 D:54 H:10 M:42 S:29
Benefit of the doubt 3J 3.0 9 312 0.502345970955 D:54 H:12 M:23 S:32
Benefit of the doubt 3K 3.0 10 312 0.503178772288 D:54 H:13 M:35 S:47
Benefit of the doubt 3L 3.0 15 312 0.505149702822 D:54 H:16 M:26 S:47
Benefit of the doubt 3M 3.0 20 312 0.505838870234 D:54 H:17 M:26 S:34
Benefit of the doubt 3N 3.0 35 312 0.506435273957 D:54 H:18 M:18 S:19
Benefit of the doubt 3O 3.0 50 312 0.506582695854 D:54 H:18 M:31 S:6
Benefit of the doubt 3P 3.0 75 312 0.506661376214 D:54 H:18 M:37 S:56
Benefit of the doubt 3Q 3.0 100 312 0.50668891286 D:54 H:18 M:40 S:19
Benefit of the doubt 4A 1.0 1 312 0.350003731381 D:45 H:8 M:6 S:19
Benefit of the doubt 4B 1.0 2 312 0.461951540827 D:52 H:1 M:58 S:54
Benefit of the doubt 4C 1.0 3 312 0.574500585795 D:58 H:20 M:43 S:40
Benefit of the doubt 4D 1.0 5 312 0.794947341552 D:72 H:3 M:29 S:37
Benefit of the doubt 4E 1.0 8 312 1.0 D:84 H:12 M:0 S:0
Benefit of the doubt 4F 1.0 13 312 1.0 D:84 H:12 M:0 S:0
Benefit of the doubt 4G 1.0 21 312 1.0 D:84 H:12 M:0 S:0
Benefit of the doubt 4H 1.0 34 312 1.0 D:84 H:12 M:0 S:0
Benefit of the doubt 5A 5.0 1 312 0.204389785065 D:36 H:13 M:32 S:51
Benefit of the doubt 5B 5.0 2 312 0.168985030714 D:34 H:10 M:21 S:8
Benefit of the doubt 5C 5.0 3 312 0.134069346351 D:32 H:7 M:51 S:51
Benefit of the doubt 5D 5.0 5 312 0.0658538250487 D:28 H:5 M:13 S:28
Benefit of the doubt 5E 5.0 8 312 0.0 D:24 H:6 M:0 S:0
Benefit of the doubt 5F 5.0 13 312 0.0 D:24 H:6 M:0 S:0
Benefit of the doubt 5G 5.0 21 312 0.0 D:24 H:6 M:0 S:0
Benefit of the doubt 5H 5.0 34 312 0.0 D:24 H:6 M:0 S:0
LT and ratings go together A 1.0 21 312 1.0 D:84 H:12 M:0 S:0
LT and ratings go together B 1.5 21 312 0.885327825354 D:77 H:14 M:11 S:2
LT and ratings go together C 2.0 21 312 0.763612882335 D:70 H:6 M:11 S:3
LT and ratings go together D 2.5 21 312 0.636436500714 D:62 H:14 M:17 S:13
LT and ratings go together E 3.0 21 312 0.5059212216 D:54 H:17 M:33 S:43
LT and ratings go together F 3.5 21 312 0.374556191754 D:46 H:19 M:36 S:29
LT and ratings go together G 4.0 21 312 0.244928310675 D:39 H:0 M:9 S:58
LT and ratings go together H 4.5 21 312 0.119426582887 D:31 H:10 M:41 S:27
LT and ratings go together I 5.0 21 312 0.0 D:24 H:6 M:0 S:0
Strong bad A 1.1 4 312 0.671263622882 D:64 H:16 M:38 S:49
Strong bad A longer 1.1 4 1200 0.702435159569 D:66 H:13 M:43 S:16
Strong bad A more votes 1.1 21 312 0.977684726689 D:83 H:3 M:43 S:55
Strong bad A longer + votes 1.1 21 1200 0.980375709149 D:83 H:7 M:37 S:23

CVGM has almost 30k active songs now, so a base time of ~1 month is pretty good (the time at Necta is even longer I believe) and up to about ~84 days for really bad songs, which we don't have many of. The whole point of this, is to explore you, as listeners, to open your horizins to new music, fnd new stuff, and find new favourites

Let the flaming begin

And of course, it arsed up my nice, formatted sheet. hehe

You know what to fix next.

How are we supposed to spam Atari ST songs and NYAN CAT theme song when goat is online now?

For example song that nobody cares about (aside from its uploader and one bored soul with 4 stars) :

Benefit of the doubt 1C 4.5 2 312 0.204389785065 D:36 H:13 M:32 S:51

vs widely respected hit:

Benefit of the doubt 1L 4.5 15 312 0.119108237789 D:31 H:10 M:13 S:49

gives only 5 days of difference over 36 days. I think the difference is too small considering compared
quality of the uploads.

Cheers,

Perhaps tonight after I get home from running all over the world doing stuff with the kids, i'll re-compile the test data list with real songs from the DB, with real numbers for times, and see how it gets altered for that. I'd like to fine tune the algo for everyone to be happy, but still keep at least a 2 week minimum with a decent spread for the songs.

Real stats data sounds great. No rush with that, though. We can observe for a bit what happens with current weights.

Btw, if you do the stats, it would be great if you could export it with something like

ID;song length;avg vote;vote count

I am really curious about some stat curves.

Cheers,
Le Goat

Then it is better that all songs get 1 month of locking time or else you will be hearing others favourite songs very often.

Is there really a problem as it is right now? I mean with people queueing "crap"?

I would be happy without the whole voting system, and just keeping the favourites option.

Yes Steve, I was thinking about that too. While having high voted songs to be unlocked sooner is nice,
these songs will get even more votes and the circle closes.

I think FG tries to prevent situation when someone with relatively long fav list can queue his favs
"ad infinitum", because they unlock faster than user can request it.

,
Le Goat

I thought the point was to keep people from trolling by queuing crap?

I think that's already handled by the quality-choke on individual sets, Deathy Like if I try to queue 3 songs in a row that have <3.0 it usually stops me.

And by 'me', I mean 'some trollish person', obviously...

Hence forth, you still need the vote system for the wider collective to decide whats crap or not

"I thought the point was to keep people from trolling by queuing crap? "

I have just requested five atari songs, so apparently that's not it.

OK, technically speaking I have experienced crap flood on CVGM only once since I started listening to the stream.
I mean very long songs with very low ratings were requested one after another on a purpose of
.

And I think it would be much easier to use "user mute" function to prevent that kind of situations.
Trying to lock all low rated songs (so that nobody can troll with a request e.g. 50 crap songs) is imho
impossible with 30k song database ("there is enough crap to flow the sewer" -debo).

Le G.F.

No, you only get to know what others like or not... and not if the music is crap or not.

The line between Good and Bad music has always been blurred, and this will continue to be a heated, drawn out discussion for years to come. It got quite nasty on Necta for a while too.

Some music is good, and some is crap, its the way it is. However, each person's tastes are unique, and so to them, a perfectly masterpiece might sound crap. Its like saying all of the Justin Bieber music is great, legendary tracks. To your average person, it's just shit on the radio, but try to tell that to a belieber hehe

The point of the voting system is not to necessarily judge the music on how technically well it is, but more for people to vote in their personal preference to how they felt it sound. It's the only data really the station can use to determine which of the better tracks should play more often, and which to leave in the dust.

I still dont feel any need for voting... the suggestion i gave earlier is better, that is lock all songs for 1 month. And maybe limit the request time total to 1 hour to avoid trolling.

To be honest i think you should spend time to fix multiple artists when uploading and possibility for image selection when uploading (also multiple option).

I have reduced the starting lock time to 15 days to keep people from complaining for the time being

See? You're too soft. When people complain that the lock time is too long, double it - and dare them to complain again!

Yeah, i should be like "screw you guys" but I try too much to keep balance. Ill change it back in a few days after tweaking the other code some more