Dealing with Queue Rot in VS 3.0

One of the things we're struggling with in the new version, is what to do about the queue.



Our initial idea is to stop automatic SiftBot discards and allow posts to stay in the queue for as long as the member wants. With this persistance though, comes the concept that the longer a post is in the queue, the harder it becomes to escape:





* Under 4 days: 10 votes to publish

* Under 8 days: 15 votes to publish

* Under 12 days: 20 votes to publish

* Under 16 days: 25 votes to publish

* Any older: 30 votes to publish




Queue limits still apply, so if you reach your quota, and you want to post something new, you would have to *discard one of your older sifts that is still in the queue. There would be no SiftBot sweeps to auto-discard older queued videos.



Any thoughts, ideas or alternatives to this system are most welcome.
twiddles says...

How would that affect the list of top 15 queued sift videos expiring soon? I kind of like the SiftBot discards, but think 4 days is not enough. I wouldn't mind seeing videos automatically saved and sent to the top of the queue for one more try.

Edit: Provided they have at least 5 votes

gorgonheap says...

Interesting concept but if your anything like me I don't want more then 4 days in the queue. I'm ready to submit some fresh stuff after it gets published/discarded. I think it's a neat concept and a *quality one at that. But I can always discard after 4 days if I want t new slot in the queue.

joedirt says...

I think this is 100% wrong.
The best idea is a no vote button, so you can remove videos from the queue, or at least a *NEW* marker so I can hide all the queued videos I've seen.

You definately do not want a longer queue with even more stuff to sift through. Imagine you were gone a week and now you have to look through 1,000 queued just to find anything worthwhile.

My idea (and this generally works manually with a save) is to re-queue good videos. So on day 3 if a video has 6 votes, do a auto-save and plop it back on top of the sift. (or day 2 or whatever) So you would auto discard lame videos (less to look at) and auto encourage good videos (much like a real sifter where larger stuff comes to top)..

No video that got less than 10 votes in 4 days... will EVER get 25 votes in 300 days. Not to mention the tail end of the older videos in teh queue never get upvotes. Most people only check out page one of the queue.

joedirt says...

Dag what you need to do is some number crunching.

Average age of video when it is promoted out of queue.
Average number of votes of videos that are discarded for lack of votes.
% of videos discarded by siftbot after 1 day, 2days, 4 days.


What you really need is a queue that only has like 20 videos in it. You would take the total number of votes for the whole queue (to scale for time of day and traffic) and calculate a videos votes/hour. A good video will always receive X votes per number of queue reviewers.

The way to avoid gaming the system is to show a unique queue (or 10 different ones) so that your friends wouldn't even be reviewing the same videos. So it would be hard to pile on votes.

lucky760 says...

>>How would that affect the list of top 15 queued sift videos expiring soon?

Instead of top expiring soon, the sidebar is "Top Save-Worthy Videos" which are a few days old with lots of votes.

>>P.S. Will it ever get discarded? i.e. a person joins and never comes back. It will sit there for eternity.

Yes, and that does bring up an interesting point. We don't want hit-and-run sifters to sign up, submit, and disappear, so maybe we continue to do auto-discards for proby member videos only?

>>You definately do not want a longer queue with even more stuff to sift through. Imagine you were gone a week and now you have to look through 1,000 queued just to find anything worthwhile.

In practice, it shouldn't work that way. If we were to allow old videos to stay queued and submitters to continue right on submitting after a few days (which is another alternative that has been floated around that I'm against), that would be the situation, but with all we active users who want to feverishly continue getting vids published, we'll decide on our own when to discard and try it again or with something else altogether.

>>My idea (and this generally works manually with a save) is to re-queue good videos.

Interesting thought, but this will unfortunately not help at all with queue rot. The so-called "lame" videos that would get discarded could very well be far from lame and just not enough of a chance at being viewed.

Perhaps we can have our suggested system running except with the "harder to escape" rule. IOW, keep it a limit of 10 and just never auto-discard, leaving it up to the user to decide when to discard. In addition to that, maybe we should not have the queue auto-sorted by most recent submissions. It would probably be better to just always sort it by fewest votes or oldest submit by default?

It's a tricky situation, this'n.

dotdude says...

What about a possible stats grid that reflects the number of the vids by channel in the Queue. OK, so there's a bit of overlap.

If the Sifter notices there are a slew of vids in a channel he/she likes, then it may speed what happens in that channel. Or maybe the Sifter is pressed for time. He/she sees a small number of vids in a favorite channel. "Oh goody, I can watch those quickly."

Most sifters have several channels they like. This may help them figure out where they want to go faster.

With the merging of channels and collectives, there's a lot choice to view. And I'm sure we'll be adding other niches as the needs arrive.

My two cents at 1:48:48 AM CST. % )

joedirt says...

You guys are nuts. really. I'm sorry, but let's say onetime tuber-fan submits something lame. Let's say they are not even "P" but a one or two comment, one or two video person. They submit a video (which is crap) and then never visit the site again.

We will have 1000s of crappy videos in the queue masking any "gems" in the rough.

So the motivation seems to be allowing medium quality or unique videos to hang around longer. So why not allow a re-queueing of videos after X number of days.


So if you really want to hang in there with your video, after day 4 you can put back on top of the queue again. After day 10 if you took no action and did not requeue, it gets deleted. Maybe others can requeue vids also. This allows you to up the escape limit, and provide a way to keep your video around "forever" if you want to, but inactive users and hit and run videos will die naturally.

jonny says...

joedirt has hit the nail on the head - the problem of queue rot is the queue getting too big. As looris (I think) pointed out in another sifttalk, vids tend to get published either within the first day or so, or while in the top 15 expiring soon list. Basically, vids in the middle of the queue are doing little more than biding time. (This is anecdotal evidence of my own experience, but I'd guess a histogram of # votes cast vs age of vid would bear that out.) You're right too lucky, default viewing of the queue by most recent submission exacerbates the problem. What you want is a queue which encourages, if not enforces, viewing of lesser seen vids - a priority queue, where each user can persistently set the weight of different variables, like age, # votes, collective entry, randomness, etc. (but of course it has to be ultra simple and quick to set up ). The default set of priorities might focus on lesser seen and/or newer vids (is it possible to track that indepedently of voting?). Again, joe has it - if you have a no-vote/not interested button, it would allow the user to remove large numbers of vids they wouldn't have voted on anyway, and make it more likely for them to see vids they might have otherwise missed and been interested in.

dag says...

Comment hidden because you are ignoring dag. (show it anyway)

JD and jonny, you bring up some good-points about the persistence of mediocre sifts - this definitely is something we need to think through a bit more.

The underlying issue is that we are structured very differently from other voting sites like Digg or Reddit. Those sites aren't as concerned with duplicate posts, or spammage - those posts just don't rise out of their deluge of submissions.

The Sift though, has developed (kind of organically) the idea that any live, non-discarded post has a seal of quality. So we don't support a shiteload of old posts that never made it anywhere.

It almost makes me wish there was some "other place" that old posts could go, to unclog the queue, but without discarding them. Maybe it's a semantics problem. After all, the discarded posts are still live- they just never made out of the queue. Perhaps "discarded" is too negative. Do we need a *library or *file invocation to shunt the old videos "somewhere" for safekeeping?

joedirt says...

I agree with that. If discarded were still 'live' somewhere, it would also prevent duplicate submissions. Like, all the I submitted that but it never made it out of the queue, good luck on your submission.

The only ones that should be out of service is dead videos, and even those should show up for reference when you submit just as an FYI.

Try "Sifted" and "Unsifted" where sifted is the queue escaped vids. Even change queue into Sifting Area, SIFTER, siftbox (get it, siftbot hangs out in the siftbox)

But, I'm not sure what you mean. If there wasn't a queue auto discarding tons of crap, I'm not sure what you mean by all live videos have a seal of quality. if they don't make it out of the queue, they sift in the unsifted, or crappy area (and maybe can still collect votes over time to finally escape).

But the incoming, voting, queue, sifting area of the site HAS TO BE SMALL AND SEPERATE from the "didn't escape the queue" area.

lucky760 says...

>>What you want is a queue which encourages, if not enforces, viewing of lesser seen vids

This is a very good idea. We are tracking video views in v3, so why don't we just default the queue to display by the least viewed videos and then by oldest submission? This way videos nearer discardation that haven't been viewed very much will always have highest priority.

What do you think of that? I think it's a perfect solution.

gorgonheap says...

Since we touched lightly on the subject will there be a way in VS3 to fix discarded videos? I have a lot of discards that I can get embeds for but of course when a video is discarded you can't alter it in anyway. This leads to multiple posts, even from the same user, of the same material. So a user could be submitting a video that is on four other accounts but not a one of them works, nor could a user get it to work if they wanted to.

lucky760 says...

That problem won't exist for future submissions as dead videos will go to the deadpool. Any videos that are discarded will only be for reasons other than a dead embed. As for past discards, no, there'll be no way to edit them.

Zifnab says...

>>So if you really want to hang in there with your video, after day 4 you can put back on top of the queue again. After day 10 if you took no action and did not requeue, it gets deleted.
I quite like this idea of joedirt's.

Another thought, I'd really like to be able to sort through the discarded video's, specifically I'd like to be able to sort them by vote count. Searching for saveworthy videos can sometimes be difficult (even with the great playlists for savable vids some members have set up). If we could list and sort the discarded videos it would be easier to find saveworthy vids. I'm not sure we can/should differentiate videos that were manually discarded or discarded by siftbot after expiring.

lucky760 says...

That could work.. If everyone could invoke *requeue on their own queued videos, but no more frequently than once every, say, two days for a given video... Hmm. Maybe that coupled with displaying least viewed videos first... I think those'd really do a great job combatting queue rot, maybe even doing away with it altogether.

twiddles says...

This is a very good idea. We are tracking video views in v3, so why don't we just default the queue to display by the least viewed videos and then by oldest submission? This way videos nearer discardation that haven't been viewed very much will always have highest priority.

What do you think of that? I think it's a perfect solution.


Will there still be "saves"? If so what happens to them? Wouldn't you have to reset their view count and/or submission date in order for them to appear near the top?


lucky760 says...

>>Will there still be "saves"? If so what happens to them? Wouldn't you have to reset their view count and/or submission date in order for them to appear near the top?

Hmm. Well, a save could still give a video another 4 day shot in the queue, but no it wouldn't appear at the default top of the queue. Still the most un-viewed videos would appear there.

grspec says...

dag uttered:
It almost makes me wish there was some "other place" that old posts could go, to unclog the queue, but without discarding them. Maybe it's a semantics problem. After all, the discarded posts are still live- they just never made out of the queue. Perhaps "discarded" is too negative. Do we need a *library or *file invocation to shunt the old videos "somewhere" for safekeeping?


I think you have just described the perfect new place for videos out of the queue but never published... purgatory.

lucky760 says...

Another evolution of my idea, in its entirety:

- Any user can *requeue his/her own video once every two days
- If a video reaches the 4 day mark, it's auto-discarded
- Queue defaults to displaying videos by a votes-to-views ratio, so videos that have more views, but fewer votes will be further back, while videos with higher votes and fewer views will be more toward the front, regardless of when it was submitted or last queued so it's based solely on video quality

Zifnab says...

>>I think you have just described the perfect new place for videos out of the queue but never published... purgatory.

I think it would be more like an Oubliette.

@lucky760
I like the sound of that. Should there be a limit on the number of times a video can be *requeued though?

lucky760 says...

>>Should there be a limit on the number of times a video can be *requeued though?

I don't think there needs to be. If someone wants to try and try and try again to get their video published, so be it. If it just isn't making it, that should tell them something about the quality of the video. And as a benefit of how this new system would work, because it would probably have lots of views from being submitted many times, it will probably linger somewhere toward the back.

twiddles says...

- Any user can *requeue his/her own video once every two days
- If a video reaches the 4 day mark, it's auto-discarded
- Queue defaults to displaying videos by a votes-to-views ratio, so videos that have more views, but fewer votes will be further back, while videos with higher votes and fewer views will be more toward the front, regardless of when it was submitted or last queued so it's based solely on video quality


If I may, I am not sure I see a need to give users a *requeue command if you will be scoring on a votes-to-views ratio. Wouldn't it be better if they were auto-discarded if their ratio was below a certain threshold for the number of days in the queue without a time limit? That way videos would stick around if they didn't receive many views and if a video is crap it will be quickly discarded after a number of views without any votes. The key would be figuring out what the thresholds should be.

lucky760 says...

That sounds very good in theory and I appreciate the input, but in practice people will often probably like to just re-post an expired queued video for another shot. That being the case, why make people start over from scratch? Their previous votes should be carried over. It is the exact same post after all.

joedirt says...

Lucky:
Queue defaults to displaying videos by a votes-to-views ratio, so videos that have more views, but fewer votes will be further back,

What does this mean?? Are you tracking javascript for play clicks? Or just for example say you are displaying 30 videos at a time, are you assuming and upvotes on any of those videos on that page applies to having seen all 30?

I personally upvote quite a few videos without ever watching them. Like based on the title you know what some are, and maybe have already seen them, or not really interested in watching them (like "WHAM- Wake Me Up")

How is this "viewed" being calculated, and how do new submissions factor in (divide by zero)? Are newly submitted vids always on the top?

I might also add into your formula votes per hour, as that will also keep hot videos at the top of the queue.

dag says...

Comment hidden because you are ignoring dag. (show it anyway)

It will be a bit clearer once the beta starts, but we can track a "views" as when the embed code is loaded. We couldn't do this before - because many people would watch the videos from the front page - we now have a solution to this.

joedirt says...

So again, are you given a queue or front page with 10 or 30 videos? Or must you click on a title to ajax load up the embed?

If you view a page of 30 videos and all 30 embeds load up, are those all considered viewed? (How does this work with FLashBlock? Because I believe from serverside, you might have that as loaded, but in the webbrowser it was no loaded at all.

twiddles says...

Lucky:

I guess I misunderstood the first time. I assume now that you mean a user could *requeue after the post had been discarded and would keep its previous votes and views. In which case wouldn't the video more than likely be queued somewhere near the end of the queue since it didn't make it out the first time? It doesn't seem like a *requeue would be very likely to make it out on any subsequent retry. Though I can see that forcing the user to make the decision would help keep the queue from being too long.

joedirt says...

The requeue would reset the date and place it on top of the queue.


Have you thought about a blind queue? Might be interesting to try out. I know a lot of videos are artificially voted up because they have 8 votes. Maybe if people didn't sort by most popular, they might look more closely at some of the 'hidden gems'.


Deano says...

But the incoming, voting, queue, sifting area of the site HAS TO BE SMALL AND SEPERATE from the "didn't escape the queue" area.
Totally agree with this. Then a smaller queue becomes automatically easier to view.


Requeue sounds logical as well but I would keep the save invocation. It's one of the pleasures of reaching gold status isn't it? If the queue is sorted by views/age I still think the save should put the video at the top of the queue.

MINK says...

i've already hit this nail on the head:
http://www.videosift.com/talk/Dynamic-Queuing

similar to things joedirt is saying. Every video should be exposed to the same number of sifters before being allowed to rot. That is the key to removing the advantage of posting during peak time.

As for vids that are great but don't make it out of the queue... well... that's because most people like shit unfortunately. Maybe there should be some kind of benefit to getting a lot of gold star members voting on your vid (seeing as gold star members are more likely to vote responsibly and thoroughly rather than just kneejerk upvoting anything with an explosion or Jon Stewart in it)

Send this Article to a Friend



Separate multiple emails with a comma (,); limit 5 recipients






Your email has been sent successfully!

Manage this Video in Your Playlists

New Blog Posts from All Members