The Song Fight Bulletin Board

How hard would it be to get the data for the entire history of Song Fight? What I’m really curious about is the long term trends. Home recording hasn’t changed much in the last three years but it’s changed a LOT over the last 20.

I’d be totally fine with just getting a big csv file with no filtering or presentation.

I'll give it a shot and report back. Looks like there's about 12,000 songs to process. Not sure if I'll get all of them but at least a wider sample.

Yeah that'd take at least a few hours to do. Maybe even a couple of days of continuous processing? I dunno how fast your machine is processing full-song LUFSes.

It's a 4+ year old laptop running some .NET libraries and ffmpeg . But I think it's done. I updated the sheet. Had to skip a few songs because the file confused my script, and skipped the entire "It's Charging" fight because I doubt it really happened in 1987. Also flipped to focus on "End Date" because the Start Date of "39 + 1 = JB" is misleading.

Impressively, it includes a song that actually measured at +3 LUFS Integrated. (R Mosquito's Everybody Calm Down)

Hope this helps. Again, the scatter is a little misleading but it looks like:

Yeah “it’s charging” was a random one-off thing that Drew did.

And that’s cool, and looks pretty much how I expected - a really wide range that slowly converges over time. I’m a bit surprised that the average seems to have always been around -12 though. Interesting how that works out.

Anyway thanks for crunching the numbers, that’s really cool to see and informative.

That data is awesome. Reve is sure to get a kick out of making people's ears bleed in 2008...

I already have them calculated for Nur Ein, so this was simple enough googling a couple Bash commands. Not worth cleaning up more. Edit: Ok I did a little more fun with Bash and gnuplot

That is very cool to see! As is the Song Fight data. I love these! Yeah data!

Also whoa. Which song is that super quiet outlier?

https://nurein.songlander.com/titles/50 ... 364B03E3AD

"Sleep Tight" - MC Eric B

Sounds like a voice memo.

http://www.songfight.org/music/whiskey_ ... rpe_wd.mp3

Octothorpe - Whiskey Drinker

was literally recorded as a voice memo on my phone

Interesting, though a couple of the “so what” questions are:

1. Does loudness affect voting? Do songs outside the average for their fight get more or less votes? Outliers are probably almost always a result of port equipment or inexperience (rarely on purpose tho not never)

1. Does loudness affect listening length? How far into a track do listeners get? Can’t tell this from voting data though I think the server logs do have some kind of listening length for media files if I recall correctly. Wonder if the jukebox has any info like that.

Having to turn my volume up or down for a track doesn't affect my voting, but if the quietness or loudness is a symptom of graver mix issues then it might well do. I've definitely found myself not voting for tracks that are super over-compressed in the past because it can be so ear-fatiguey (to use the scientific term).

jb wrote: ↑

Mon Nov 04, 2019 9:25 am

1. Does loudness affect listening length? How far into a track do listeners get? Can’t tell this from voting data though I think the server logs do have some kind of listening length for media files if I recall correctly. Wonder if the jukebox has any info like that.

That would be interesting to one but there’s nothing on songfight.org which would be recording that; you need a custom player setup that sends telemetry back. Bandcamp has a custom player that does that for their album player; perhaps that’s what you’re thinking of?

I think at best we can maybe see how many bytes of a file were downloaded but download speeds are way faster than play speeds so that wouldn’t tell us anything interesting, and it would only work for the tiniest minority of listeners anyway.

fluffy wrote: ↑

Mon Nov 04, 2019 10:37 am

jb wrote: ↑

Mon Nov 04, 2019 9:25 am

1. Does loudness affect listening length? How far into a track do listeners get? Can’t tell this from voting data though I think the server logs do have some kind of listening length for media files if I recall correctly. Wonder if the jukebox has any info like that.

That would be interesting to one but there’s nothing on songfight.org which would be recording that; you need a custom player setup that sends telemetry back. Bandcamp has a custom player that does that for their album player; perhaps that’s what you’re thinking of?

I think at best we can maybe see how many bytes of a file were downloaded but download speeds are way faster than play speeds so that wouldn’t tell us anything interesting, and it would only work for the tiniest minority of listeners anyway.

Yeah, my main thought is that this data is fascinating but only that-- it's interesting to see that songs have basically been always the same loud, and interesting to see what the averages are-- the bell-curve seems to be a little to the right on the chart. But does it lead to a useful conclusion about us? Is there utility in this particular aspect of the catalog?

vowlvom wrote: ↑

Mon Nov 04, 2019 9:45 am

Having to turn my volume up or down for a track doesn't affect my voting, but if the quietness or loudness is a symptom of graver mix issues then it might well do. I've definitely found myself not voting for tracks that are super over-compressed in the past because it can be so ear-fatiguey (to use the scientific term).

If a particular artist consistently turns in a super quiet or a super loud mix I tend to have a preconceived negative feeling toward their submissions. I consciously try to overcome it, and I'm fairly proud that I realized it (if only I could see all of my biases so easily) but it's still something I have to pay attention to. All that to say, I'm pretty sure that if one of these outliers are on the fence of a vote it could be the straw that capsized the boat.

Seeing this data, though, I'm now tempted to use a louder mix on songs that I think have a chance of placing well with the songfight voters. To be fair, that's not many songs I submit and typically the ones I thought would do well don't and a couple have surprised me with doing better than I thought they would.

Loosing votes here and there because of loudness doesn't really matter since people are still friend flooding the votes anyhow...

ujnhunter wrote: ↑

Tue Nov 05, 2019 9:50 am

Loosing votes here and there because of loudness doesn't really matter since people are still friend flooding the votes anyhow...

Heh, this is true. Though if you approach it as practice-- practice songwriting, practice performing, practice producing --you want all the feedback you can get about how to get better, and apparently loudness might be a kind of indicator or measure. Certainly professional mixers have a process and methods that depend on their ultimate goals. And that's true for Song Fight as it is for anywhere. If you wanna win the live fight at SFL it sure helps to be as many of these as possible: hilarious, cute, local.