It seems that a recent upload from Admiral Twin "You Hung the Moon" has revealed a weird behaviour of the algorithm used to post songs. It now has 774 bumps, it was uploaded more than 4 days ago, and it is still not posted. I understand that songs from popular artists require more bumps, but a song by HBA just got posted with less than 500 points. I also understand that a song by AT (In my veins) was also revived, but that was more than two days ago, which I believe is the limit.

Is this a bug, or is there something I really don't understand in the logic behind posting?

This is something obiously that the admins should answer.

My guess, without actually looking, is that it is the new HBA, which was uploaded within the last 24 hours.

It makes it a hotter track because there has been a lot of activity in a shorter time frame than the Admiral Twin song.

From what I understand:
A song can't post within 48 hours of a revive or posting of another song from the same band, no matter how many bumps it has. And then, even after that, there may still be a queue system for genres and other songs that have "qualified" to post, since the fastest cadence of song posting looks to be one every 30 min. If the queue is stacked high, a track could be delayed significantly beyond 48 hours, even if it's hot to trot.

How high and how long a track gets on the Hot page has to do, I believe, mostly with the rate of bumps once it posts, So if a song gets 2000 bumps before it posts, and 10 after, it won't live nearly as long on the hot page as a song that got 200 bumps before posting and 1800 after.

All that being said, I agree that there's something broken about how revives interfere with popular new tracks. There have been multiple occasions like AT's where an upload was quickly followed by a Revive of a song from the same artist that prevented the upload from getting posted and the popularity it should have. This is actually an example of the site _mechanics_ interfering with the site _intent_. One person's vote (the Revive) can effectively and accidentally torpedo the votes and love of many, and misrepresent the popularity of a given artist and/or track.

I'm thinking it had to have been a combination of things that piled up. The song was sandwiched between two revives, but it did have at least a day as I recall after the first 2 day revive effect had passed when it was very hot - up to over 650 within 24 hours but it still didn't post. Things slowed some after the second revive blocked it, but it still got over 100 more. But because of having to wait for two more days for another revive, I think the time frame has gotten falsely stretched out so the hotness appears to be lessened. Still...there are songs that don't have that many points that have been up for months and they get in. It has to be some kinda fluke that several unfortunate circumstances all converged at once to let them fall through a big crack. It just seemed like this song was obviously hot and should have been on the list, but something prevented it. Don't know if it's a bug or just an abnormality, but I agree with iyzie that the current revive system can cause misrepresentations to happen.

our system looks at a song and draws from a wide array of user data to determine the "diversity" of bumpers. this is the same system that we use to prevent bigger bands from simply flashmobbing thesixtyone by bringing their massive thousand-man fan armies from myspace.

Thanks James, I understand some of the logic. But is it because 204 unique bumpers is not considered diverse enough, or because its pretty much the same 204 bumpers that bumped other Admiral Twin songs?

Just trying to understand here.

Good question, carpex. I would definitely like to understand why this is happening.

I'd like to clarify what the problem is here...

Are you saying that this song is suspected (by the algorithm) of abusing the system?

Do you agree with that assessment upon human inspection?

Will there be any restorative action to get it "unstuck in time"?

How can a repeat of this problem be avoided?

The activity page for classic pop shows You Hung The Moon with 791 points in 5 days listed on the second page. It is below ten others with much lower points in many more days. This surely indicates a bug, yes?

New Theory on You Hung The Moon:
Classic Pop has a queue with low priority relative to other genres, or else some of the genre-specific queues are unbalanced are broken.

Potential evidence:
Sorting by bumps on Classic pop shows You Hung the Moon as #3 all time, but still unposted. The only other unposted song on page 1 there is Glen Case's Summer Breeze, and it's 26 days old. Page 2 has 3 more unposted songs, 2 with enough bumps to post under most circumstances, and both 2-3 weeks old.

The most recent Classic Pop song to post is: Lucky Soul 19 days ago.

Conclusion: I think perhaps some of the genre posting queues are unbalanced or broken.

The reason this is a serious site bug: If the posting rules are unbalanced based on genre, then an artist may unwittingly sabotage themselves by choosing one genre over another equally fitting one, and listeners' voices/bumps count for less (in a way unintended by the site) if the weighting penalizes one genre over another.

We've known for a while that different genres post differently. It was a feature added some time ago to force us to listen to country, rap, and classical music ;)

Agreed petiejoe, but I don't think any genre formula was designed to keep a 400+ bump song away from posting for 3 weeks, or a 700+ bump song off for 4 days.

Here is another to add to the underweight genre formula thought. Across the Air covered by Katie Enlow it's listed in the experimental genre and I've had a hard time for awhile trying to understand how it's still sitting there unposted.

guys, there's nothing wrong with our genre system. pls refer to my post earlier in this thread.

Informatively uninformative, james.

I must then retract my theory that genre formulas are broken.

Looking at james' post:
"our system looks at a song and draws from a wide array of user data to determine the "diversity" of bumpers. this is the same system that we use to prevent bigger bands from simply flashmobbing thesixtyone by bringing their massive thousand-man fan armies from myspace."

Leads me to think carpex must be right:
"... is it because 204 unique bumpers is not considered diverse enough, or because its pretty much the same 204 bumpers that bumped other Admiral Twin songs?"

One possible deduction is that a high overlap of individual bumpers for the songs of a given artist is detrimental to their posting and placement. i.e. the same people voting again and again for the same artist count less for that artist.

The "scaling" implication is that if Band X brought their 900 fan army to t61 and all 900 bumped every one of their songs, that army's bump power - for affecting Band X - would be significantly curtailed with subsequent new songa.

That further implies that perhaps just a few more individual bumpers who are not currently big AT followers would be enough to push them over the top.

These are my best speculations from the evidence presented.

Well, pardon my exasperation, but that is a very arbitrary decision by "the system" to consider repeat bumpers as bad bumpers.

I don't know how anyone could look at the bumper list on this song and not see diversity. There are listeners at all levels and bumps from 1 to 20. Many of the bumpers were here long before AT joined. Sure, they are AT fans now. They are also Temple Scene fans, Glenn Case fans, Sumkid fans, they are fans of diverse bands. Is that in the wide array of data you checked?

I don't know what metadata you've told the system to check, but it is returning false positives and that is a bug.

I applaud your desire to block overinfluence by superbands, but this is clearly not the case here.

The other possibility is that the line about excessive influence by thousand man armies (which iyzie has decimated to 900 ;) ) is a red herring, and the intent is just to raise the bar for bands that have developed a repeat fan base here.

Is that really what you want to do? Tell 213 listeners with 843 bumps that their votes don't count? Talk about your hanging chads...

Lots of bumps and lots of listeners is pretty much the essence of hot. It's not just the list of things that are hot that don't have fans yet.

I'm beginning to think this is a Catch 22. You can get new listeners by getting to the homepage but you can't get to the homepage because you don't have new listeners. And as soon as you get a new listener they aren't new any more so they don't count the next time.

I don't care how you slice it, this just isn't right. Can you tell us what was wrong in a little less vague language, please. If you really don't want a certain behavior to repeat it is only reasonable that you let us know what it was we did that you didn't like.

@utterlylinda

One thing to note - it really is a speculation on your part about "the system", and devaluing the *bumps* of high level or repeat bumpers of a high level band is certainly not arbitrary if one of the stated goals is to raise the bar for more popular artists.

The most confounding thing to me is that this particular song seems to be in a unique situation, which implies either something is broken or we are missing something. We've been assured nothing is broken, thus we are left with us missing something. That something may be an algorithm we disapprove of, but it is simply not transparent to us.

Personally, I agree that "Lots of bumps and lots of listeners" is the essence of Hot, and we have seen that there is a significant diversity in the unique bumpers between different AT songs. This may still be below some numerical threshold for the cold equations of posting. There may still be weighting and queue issues (while not broken per se, they might still be unbalanced from my perspective - if they in fact exist) due to the genre of music for this particular song.

On the other hand, the song may be sitting there 98% of the way, just needing a few more bumps from a few new people to kick it over. Shouldn't be too hard to manage that, eh? Time to fire up the tunebox, I think.

Yep, it is speculation on my part for sure, but barring any actual stated goals that's really my only recourse. If there are some, please let me know.

All I know (or have been told) is the system determines the diversity of bumpers. It was kind of implied that this was to prevent an influx of evil hordes from bands trying to subvert the system. However, since that wasn't the case here but still no acknowledgment that things weren't working as planned came about, I must speculate again that there is something going on that we don't understand, but they do.

i.e., your missing info theory does seem like the answer. Well...the answer to defining what what we need to know anyway. Knowing that we don't know something is at least knowing something. Egads, I sound like Rumsfeld!

Perhaps it would be instructive to revisit a couple of questions (for the developers) I asked earlier:

Are you saying that this song is suspected (by the algorithm) of abusing the system?

If so, Do you agree with that assessment upon human inspection?

If not, what trait is the algorithm focusing on that causes this to be such a unique case?

i'm obviously not at liberty to disclose exactly how it works, though i wanted to emphasize that the system isn't "punishing" the song. if you're familiar with the concept of "normalization," we basically draw from a wide array of behavioral data/patterns to understand the song's universal appeal in the overall community. this is simply a measure to encourage a level playing field between more established acts on thesixtyone (or popular groups coming into it) and newer ones.

James, I agree with the principles. I also don't have any particular attachments to this song. I just don't understand how the principles apply in this case. How can one look at the data of this song and decide that it has less "universal appeal" in the community than other songs that are posted?

In other words, the law doesn't seem to match the justice.

I'm gonna go out on a whim here, and say that any system that nets a wide variety of data in judging whether a song should be created is going to create some anomalies that will slip through the loopholes then fall through the cracks.

"Hmmm, that song has gotten a ton of bumps, spiked quickly and has been bumped just a few times by a lot of users interested in "folk" music, but the song itself is "alternative" and few users under level four who lack original avatars and use Linux have bumped it. So let's give it another week."