The Halloween Competition sweep

[MythicMind]

Mine was swept, going from—I'm not kidding—a steady 4.91 for three weeks with about 140 votes down to 4.83.

 

[AwkwardMD]

Mine was swept, going from—I'm not kidding—a steady 4.91 for three weeks with about 140 votes down to 4.83.

Not all fraudulent votes are 1s.

 

[FrancesScott]

https://en.wikipedia.org/wiki/Click_path

 

[AwkwardMD]

For those of you playing the home game, it would require a score of 4.83 with 70 votes, followed by 70 fraudulent 5 star votes to move a score from 4.83 to 4.91 at 140 total votes.

 

[Actingup]

Everything I've learned about sweeps in the last ten years is that they go through the entire database of stories - certainly the last ten years' worth. I still see on occasion the scores on my very first stories lose a vote or two, and the score bumps up a bit.

Just about every time there's a contest sweep, I see an adjustment or two in my scores - I can't remember the last time I entered a contest.

Apologies for being picky, but that would also work if they are sweeping the votes recorded in the previous X months, whether those votes were on a new story or an old story, right? Otherwise they would be using a lot of processing power to sweep voting data that's already been swept.

 

[MythicMind]

Not all fraudulent votes are 1s.

There would have to be dozens of "fraudulent" 5-star votes. Where would those come from? And determined how?

 

[Bramblethorn]

There would have to be dozens of "fraudulent" 5-star votes. Where would those come from? And determined how?

There are some technical clues that might be used to detect spurious voting, but in previous discussions the site owners have requested that people not discuss these since it might let trolls figure out how to dodge the sweeps. Posts ignoring that request are likely to get the thread locked.

 

[IWroteThis]

There would have to be dozens of "fraudulent" 5-star votes. Where would those come from? And determined how?

I would guess it's the exact same process as how they determine that 1-star votes are fraudulent? The rating itself doesn't prove anything.

 

[MythicMind]

Just seems unlikely that the story would receive a steady stream of somehow-fake 5-star reviews, enough to keep it steady between 4.9 and 4.92 for at least two weeks. It wasn't like it was originally at 4.83 and then it suddenly popped up. And if my random story was being fed a steady stream of fake 5-star reviews, why didn't the rest of the stories? It just doesn't sit right with me, is all I'm saying.

 

[AwkwardMD]

Just seems unlikely that the story would receive a steady stream of somehow-fake 5-star reviews, enough to keep it steady between 4.9 and 4.92 for at least two weeks. It wasn't like it was originally at 4.83 and then it suddenly popped up. And if my random story was being fed a steady stream of fake 5-star reviews, why didn't the rest of the stories? It just doesn't sit right with me, is all I'm saying.

All I'm telling you is the ratio. It takes an increasingly, and eventually exponentially, higher number of 5 star votes to move the needle above 4.8.

 

[IWroteThis]

Just seems unlikely that the story would receive a steady stream of somehow-fake 5-star reviews, enough to keep it steady between 4.9 and 4.92 for at least two weeks. It wasn't like it was originally at 4.83 and then it suddenly popped up. And if my random story was being fed a steady stream of fake 5-star reviews, why didn't the rest of the stories? It just doesn't sit right with me, is all I'm saying.

And I'm not saying they were all fraudulent any more than I would foolishly claim that every 1-star vote removed is fraudulent.

My point was, it's the same process, so either you support the sweeps and their attempt to remove fraudulent ratings or you don't. You can't pick and choose when you support it. (If you are universally opposed to the sweeps and I just missed that, my apologies.)

 

[MythicMind]

All I'm telling you is the ratio. It takes an increasingly, and eventually exponentially, higher number of 5 star votes to move the needle above 4.8.

Yes, IF the score starts at 4.8. But this story did not start there...it never went below 4.9 until the purge.

 

[MythicMind]

And I'm not saying they were all fraudulent any more than I would foolishly claim that every 1-star vote removed is fraudulent.

My point was, it's the same process, so either you support the sweeps and their attempt to remove fraudulent ratings or you don't. You can't pick and choose when you support it. (If you are universally opposed to the sweeps and I just missed that, my apologies.)

I guess it doesn't matter, since no one can say how it even works. It just makes me much less enthusiastic about trying for another contest. Either because the sweeps aren't accurate/neutral or stories get bombed with enough fake scores that it skews the voting process.

 

[MythicMind]

I will say, readership seems to make a difference. Of the three winners, they have 512, 892, and 2220 followers.

 

[EmilyMiller]

Either because the sweeps aren't accurate/neutral or stories get bombed with enough fake scores that it skews the voting process.

Ya think?

Enter comps to have fun and maybe reach some readers you wouldn’t otherwise. Get the stupid ideas that they are a level playing field and measure merit out of your head before they drive you crazy.

It’s a bit of fluff - it signifies nothing.

 

[ElectricBlue]

Apologies for being picky, but that would also work if they are sweeping the votes recorded in the previous X months, whether those votes were on a new story or an old story, right? Otherwise they would be using a lot of processing power to sweep voting data that's already been swept.

My observation is that all stories get swept, rather than only more recent stories, which is what a lot of people seem to think. My guess is that each sweep scrutinises votes recorded since the previous event.

Although, several years ago there was evidence of several very deep sweeps in a short period using different criteria, which was a bit like one of Stalin's purges in the later 1930s. A bunch of people got really excited in that one, because huge numbers of votes disappeared and story rankings shifted all over the place.

 

[TheLobster]

All I'm telling you is the ratio. It takes an increasingly, and eventually exponentially, higher number of 5 star votes to move the needle above 4.8.

<pedant-mode type="math"> It's akshually hyperbolic, not exponential. </pedant-mode>

 

[RejectReality]

Mine was swept, going from—I'm not kidding—a steady 4.91 for three weeks with about 140 votes down to 4.83.

If your score goes down when sweeps run, you have overzealous fans "helping" you by multi-voting 5 on you or 1-bombing others. Unless you tell them to stop, they automatically assume approval for their shenanigans. Most of them aren't truly slimy. They see themselves as mischievous little white knights, and when told you don't want that kind of help, they cut it out. It also helps to let them know they're actually damaging your score rather than helping it when they pull that junk.

There's a reason almost my stories end with a note saying to only vote once, and most contest stories also mention not bombing other people.

****

As to the nature of the most recent contest, I was paying zero attention. Didn't even know the contest had ended. I'm automatically suspicious because so many are convinced the site is in the early stages of heat death. So, everything is going to be viewed through those shit-colored glasses, and likely overblown. But with no personally observed data points... *shrug* Add another thing to the sky is falling list and have fun, I guess.

If the sweep was in fact half-assed, then the kind of unsophisticated chicanery your average cheeky fan pulls is exactly the kind of thing that would get wiped out. They have no idea how to evade the sweeps, because most of them don't even know they exist.

 

[StillStunned]

I didn't see any visible change near the end of the contest. But I'd downloaded my stats on 28 October, and I did it again as soon as the winners were announced on 30 October. Yesterday I compared the two sets. Not a single vote had been swept, on 68 stories.

Usually this makes me glad that I'm out of the contest game, because it means that I'm not being targeted by people so petty that they want to rig the system. But I've had some suspicious drops in my scores even so, and it would have been nice to have them looked at.

 

[iwatchus]

I can think of at least one more mechanism for the sweep to work. Cookies, for example, although that one is easy to beat. Browser fingerprinting as well, but that's a bit more advanced (for Lit) and still not unbeatable

There are far more mechanisms that would be easy to implement you have not considered yet. Again, I am not going to speak of any of them here.

Vote sweeping, as it is currently implemented, is 1000% beatable.

Any mechanism is going to be flawed in some ways. And there is apparent evidence that they have been defeated. Of course, they appeared to only run a subset of the normal sweeps this go round, which means that it was easier.

Having never tried to vote fraudulently, I cannot say from first hand knowledge that they are beatable, but circumstantial evidence agrees with common sense here.

Would 1000% beatable mean that every sweep actually amplified your invalid vote ten fold? That would be very effective and encourage Manu to reduce the number of sweeps!

 

[AwkwardMD]

<pedant-mode type="math"> It's akshually hyperbolic, not exponential. </pedant-mode>

Maybe I meant logarithmic