Seemingly widespread ratings manipulation

[SkyBubble]

registration required to vote and a public voting record.

That sounds OK.

I'm not one who worries about my ratings or vote totals. I just care about reaching as many readers as possible.

 

[TheLobster]

Monday, after a sweep:

View attachment 2618248

Today:

View attachment 2618249



Surely, just a coincidence.

The same coincidence, over and over again, story after story...

Grammar Police is ever vigilant.

 

[RejectReality]

Everybody's go-to... Required registration = vote totals tanking = every troll vote having an even more oversized impact = nothing to track when they can tank you with a single 3.

Joe Schmoe isn't going to make the effort to sign up. And that's the Joe Schmoe who is the 1 in 100 who will vote when there are no hoops to jump through.

Jackass McTrollsalot will sign up for three accounts, because he's on a mission.

It's not even remotely a solution. It's a whole different and even more frustrating set of problems that comes with a side of what limited feedback we get being flushed straight down the shitter.

 

[onehitwanda]

Ten coincidences equals a pattern

But it's only statistically significant when there's 30 coinkydinks swishing their tails in a circle.

 

[AwkwardlySet]

Everybody's go-to... Required registration = vote totals tanking = every troll vote having an even more oversized impact = nothing to track when they can tank you with a single 3.

Joe Schmoe isn't going to make the effort to sign up. And that's the Joe Schmoe who is the 1 in 100 who will vote when there are no hoops to jump through.

Jackass McTrollsalot will sign up for three accounts, because he's on a mission.

It's not even remotely a solution. It's a whole different and even more frustrating set of problems that comes with a side of what limited feedback we get being flushed straight down the shitter.

But is it really a problem if we get fewer votes? It's comments that we all crave, not arbitrary grades.

And if it's gonna be fewer but more genuine votes, especially if the administration makes an effort to detect trolls who use double or tripple accounts? It would take less effort than the sweeps for sure.

 

[consistentlyrepudiated]

But is it really a problem if we get fewer votes? It's comments that we all crave, not arbitrary grades.

And if it's gonna be fewer but more genuine votes, especially if the administration makes an effort to detect trolls who use double or tripple accounts? It would take less effort than the sweeps for sure.

It's like reasoning with cult members; their leaders can never be wrong. A house infested with roaches is the only world they know.

 

[EmilyMiller]

It's like reasoning with cult members

Or people for whom the current arrangements work just fine…

 

[RejectReality]

But is it really a problem if we get fewer votes? It's comments that we all crave, not arbitrary grades.

And if it's gonna be fewer but more genuine votes, especially if the administration makes an effort to detect trolls who use double or tripple accounts? It would take less effort than the sweeps for sure.

When you reduce the overall pool of votes, you don't get quality over quantity. All you get is even greater extremes of people who love you, and people who hate you. Among the people who love you are also the majority of the trolls fucking with everyone else out of some twisted sense of white-knighthood to begin with. You get even more inflated scores that make everyone's ego feel warm and fuzzy until it reaches the point where everyone's in that tip-top pinnacle, and a single 3 or 4 can destroy your visibility.

Nobody even needs an extra sock-puppet at that point.

I watched this happen. I'm not speculating. I'm reporting the results of this terrible ego-stroking experiment elsewhere.

 

[AwkwardlySet]

When you reduce the overall pool of votes, you don't get quality over quantity. All you get is even greater extremes of people who love you, and people who hate you. Among the people who love you are also the majority of the trolls fucking with everyone else out of some twisted sense of white-knighthood to begin with. You get even more inflated scores that make everyone's ego feel warm and fuzzy until it reaches the point where everyone's in that tip-top pinnacle, and a single 3 or 4 can destroy your visibility.

What you're saying is pure conjecture. This isn't about the sample size. This is about making it considerably harder for ANYONE to vote fraudulently. And it's about making it easier for those in charge to detect fraudulent behavior.

Scores have never been about quality. You can't make people vote about how well written a story is rather than about its content.
But you can make it so that everyone gets to vote only once, whatever their intentions are. You can make it harder to cheat, and you can make it so that any user who's cheating gets caught eventually, and all of their votes can then be easily removed by deleting the fraudulent account. No server-taxing sweeps needed.

I watched this happen. I'm not speculating. I'm reporting the results of this terrible ego-stroking experiment elsewhere.

If you're gonna bring up the experience from other platforms, I don't think this argument works in your favor. I believe you mentioned something like this happening on Lush, right? Let's see how other platforms work.

SOL doesn't allow voting without an account. AO3 doesn't allow rating at all, just kudos, which are a kind of a "like" system.

If we go to non-erotica websites, it's even harder to find any place where unregistered voting is allowed. Off the top of my head, I can't recall any site that allows it. Amazon Books doesn't. Goodreads doesn't. RoyalRoad, a huge and influential website for fantasy, doesn't even let registered users vote. They first need to read sufficiently and "level up" before they are allowed to review.

You need to look outside your own experiences. If we are going to call what most other platforms do in this sense "common wisdom", then that argument goes in favor of my standpoint, not yours.

 

[EmilyMiller]

Opinions on this matter are entrenched. But the status quo crowd are likely to become the Maginot Line, circumvented by the troll Panzers (I learned all about this from a family member and WII buff recently ).

 

[RejectReality]

What you're saying is pure conjecture. This isn't about the sample size. This is about making it considerably harder for ANYONE to vote fraudulently. And it's about making it easier for those in charge to detect fraudulent behavior.

Scores have never been about quality. You can't make people vote about how well written a story is rather than about its content.
But you can make it so that everyone gets to vote only once, whatever their intentions are. You can make it harder to cheat, and you can make it so that any user who's cheating gets caught eventually, and all of their votes can then be easily removed by deleting the fraudulent account. No server-taxing sweeps needed.




If you're gonna bring up the experience from other platforms, I don't think this argument works in your favor. I believe you mentioned something like this happening on Lush, right? Let's see how other platforms work.

SOL doesn't allow voting without an account. AO3 doesn't allow rating at all, just kudos, which are a kind of a "like" system.

If we go to non-erotica websites, it's even harder to find any place where unregistered voting is allowed. Off the top of my head, I can't recall any site that allows it. Amazon Books doesn't. Goodreads doesn't. RoyalRoad, a huge and influential website for fantasy, doesn't even let registered users vote. They first need to read sufficiently and "level up" before they are allowed to review.

You need to look outside your own experiences. If we are going to call what most other platforms do in this sense "common wisdom", then that argument goes in favor of my standpoint, not yours.

So, there's no trolling on SOL, eh? LOL All of mine that come out of the gate at midnight with back-to-back 1s in minutes would beg to differ.

And the only reason the majority of scores aren't in the high 8.5s and 9s is because they're artificially deflated by the voodoo math they use to spread the scores out. That's exactly why it was implemented, because that's where the majority of scores sat, and the paid members complained about the scores being useless for selection.

SOL has always been members-only read/vote as well. It's a whole different ball of wax when you take a previously free site and try to implement controls like this. A huge majority of people come here precisely because they're not required to sign up.

Lush is a much better analog, because it had more or less the exact same setup in the beginning. The moment sign-up to vote became required, rather than a setting fragile authors could enact to protect themselves from the uneducated peasants who couldn't appreciate their brillance, those of us who allowed anyone to vote saw our vote numbers absolutely fall off a cliff. Not half. More like a quarter, at best.

 

[Bramblethorn]

What you're saying is pure conjecture. This isn't about the sample size. This is about making it considerably harder for ANYONE to vote fraudulently. And it's about making it easier for those in charge to detect fraudulent behavior.

Not at all clear that it achieves that, though.

If you reduce voting rates by 75% (taking @RejectReality's example from Lush, which seems pretty plausible), you are effectively quadrupling the voting power of every troll-created voting account, which increases the damage trolls can do before being identifiable as trolls.

Say I have a story that's sitting on 4.80 with 100 votes, and somebody wants to knock it down to 4.64 for the sake of a contest or whatever. It takes four 1* votes to do that, and when a story that's otherwise scoring highly gets four 1* votes - especially if they have the same technical fingerprints - it's pretty obvious that something shady is going on.

Now, say we restrict voting to logged-in users, and it drops voting by three quarters like @RejectReality mentioned it doing on Lush. When my story is sitting on 4.80 with only 25 votes, it only takes one 1* vote to drop it down to 4.70. How are you going to look at one single vote and determine whether it's somebody trying to rig the contest, or just somebody who didn't like my story?

It might discourage the systematic toplist downvoting, and that would be nice. But given the other consequences it feels like a last resort, and currently it's not clear whether @Laurel and @Manu have explored other options. AFAIK it's not even clear whether they have acknowledged it as a problem.

Lush is a much better analog, because it had more or less the exact same setup in the beginning. The moment sign-up to vote became required, rather than a setting fragile authors could enact to protect themselves from the uneducated peasants who couldn't appreciate their brillance, those of us who allowed anyone to vote saw our vote numbers absolutely fall off a cliff. Not half. More like a quarter, at best.

This is particularly bad because even if you could completely prevent fraudulent voting this way - which you can't - it's liable to make the scoring so noisy that top lists would be dominated more by luck than by story quality.

 

[Stanfield2025]

But is it really a problem if we get fewer votes? It's comments that we all crave, not arbitrary grades.

And if it's gonna be fewer but more genuine votes, especially if the administration makes an effort to detect trolls who use double or tripple accounts? It would take less effort than the sweeps for sure.

Comments are the best, but not everyone gets comments. I prefer having more votes, even if I wish more votes were genuine.

 

[TheLobster]

When my story is sitting on 4.80 with only 25 votes, it only takes one 1* vote to drop it down to 4.70. How are you going to look at one single vote and determine whether it's somebody trying to rig the contest, or just somebody who didn't like my story?

By looking at other votes cast by the same account. And if that 1* is its only vote, then that's a rather obvious sign in and of itself.

 

[Actingup]

Grammar Police is ever vigilant.

*are. You’ve been waiting patiently for somebody to bite, haven’t you?

On the subject at hand… I think that a radical approach could produce interesting results. Take the top lists down completely- they are invalid due to vote manipulation, but also they discourage innovation. I can’t ever beat My Sister Has Tits Chapter 37. Instead, just show a random selection of highly rated stories in each category. For those that want a Hall of Fame, create some curated lists on various themes, and recognise the legends in those.

 

[consistentlyrepudiated]

Or people for whom the current arrangements work just fine…

Or the ones for whom the rot works just fine... I’m sure many in North Korea find their arrangements comfortable... or rats in a sewer, feeling snug in the dark, certain that fresh air is a toxin.

Either way, nothing changes; only decays. Since the outcries are only going to get louder, I might as well lie back and enjoy the concert...

 

[StillStunned]

Or the ones for whom the rot works just fine... I’m sure many in North Korea find their arrangements comfortable... or rats in a sewer, feeling snug in the dark, certain that fresh air is a toxin.

Either way, nothing changes; only decays. Since the outcries are only going to get louder, I might as well lie back and enjoy the concert...

Suffered much from voting manipulation, have you? With your.... how many stories?

 

[OverconfidentSarcasm]

So, there's no trolling on SOL, eh? LOL All of mine that come out of the gate at midnight with back-to-back 1s in minutes would beg to differ.

Of course there are trolls on SOL. There are trolls everywhere. But I think you're missing the point. And I don't get where the constant bite-reflex comes from when discussing this topic. This thread is about large-scale score manipulation. And the notion that every troll is on some kind of personal crusade and willing to invest the time and energy to create hundreds of accounts, is just ridiculous. That's not how people work.

If you just open any story, hit F12, and then access the network tab when casting a vote, you'll see that, right now, there are no client-side safeguards against mass-scale bombing. And I highly doubt they implemented any server-side safeguards either.

The site submits your user agent, browser language, vote score, and your user-ID if you're logged in... and that's it. It doesn't even submit a referrer URL. So, it doesn't check how long you've been on the site, or whether you accessed the voting page by (at least) accessing the previous story page as an indicator for actually reading what you're voting on. All you have to do is open a private browser window, access the voting page of a story, and cast a vote. And you can do that as many times as you want.
They might check your IP address. But, for that to work in any meaningful way, it would require them to actually save the IP addresses of every anonymous vote for the last 24 hours for every story, and then check those IP addresses with EVERY anonymous vote coming in. Call me crazy, but I just don't believe they'd spare the resources. Especially since they are bound to know that TOR is free.

So, yes, trolls exist. But, right now, anyone with even a basic understanding of how google works can search for Voting Bots or Scripts and have them run all day long.

The goal here isn't to make fraudulent voting impossible... it's to make it no longer be THAT easy.

My personal opinion? Just implement a check for the referrer URL and how much time was spent on the site before the vote got cast... and then DON'T announce it. Unless you at least started reading the story, the vote should be ignored. It should take the few tech-savvy trolls on here a while to figure that out. And then there're even fewer tech-savvy trolls who'd be able to write a custom voting script that accommodates for the new requirements.

*EDIT*: Just tried it to make sure I'm not spouting crap. They apparently do check the IP addresses in some way. But, as I wrote, that isn't a real hindrance.

 

[Bramblethorn]

My personal opinion? Just implement a check for the referrer URL and how much time was spent on the site before the vote got cast... and then DON'T announce it. Unless you at least started reading the story, the vote should be ignored. It should take the few tech-savvy trolls on here a while to figure that out.

How would this referrer URL check behave in the following scenarios?

• Being privacy-minded, I install a browser extension which by default disables referrer URL. I visit Literotica, read some stories, and attempt to vote.
• I start reading a Lit story. It's too long to finish in one sitting, so I bookmark it and close the browser. Later I reopen it and resume reading from the bookmark, then attempt to vote.
• A friend reads a story on Lit, decides to recommend it to me, then sends me the URL via some non-browser app (Discord, email, SMS, etc.) I click that link, read the story, then attempt to vote.
• Somebody posts a bunch of Lit story recommendations on their Wordpress blog, or any other site which uses rel="noopener noreferrer" for external links (something that's often recommended for privacy + security reasons). I follow those links, read the stories, and attempt to vote.

And then there're even fewer tech-savvy trolls who'd be able to write a custom voting script that accommodates for the new requirements.

I'm not sure I share your optimism on this front. The toplist manipulation seems more likely to be a script than one person manually downvoting hundreds of stories; adding a faked referrer URL to that kind of script would be a single LoC.

(And a site like Literotica shouldn't be voluntarily increasing the amount of data that it collects on its users. That's getting into "burn down your house to kill a spider" territory.)

 

[OverconfidentSarcasm]

How would this referrer URL check behave in the following scenarios?

• Being privacy-minded, I install a browser extension which by default disables referrer URL. I visit Literotica, read some stories, and attempt to vote.
• I start reading a Lit story. It's too long to finish in one sitting, so I bookmark it and close the browser. Later I reopen it and resume reading from the bookmark, then attempt to vote.
• A friend reads a story on Lit, decides to recommend it to me, then sends me the URL via some non-browser app (Discord, email, SMS, etc.) I click that link, read the story, then attempt to vote.
• Somebody posts a bunch of Lit story recommendations on their Wordpress blog, or any other site which uses rel="noopener noreferrer" for external links (something that's often recommended for privacy + security reasons). I follow those links, read the stories, and attempt to vote.

Did you somehow miss the part where I wrote "implement a check for the referrer URL AND how much time was spent on the site before the vote got cast"? Also, as long as you visit the site, you WILL send a referrer, even if it's just "literotica.com". Because voting is a FETCH request.

Though, I also have to say, if you are THAT privacy oriented that you would intentionally disable standard browsing functions, you kinda just have to live with the fact that standard functions of websites then won't work.

I'm not sure I share your optimism on this front. The toplist manipulation seems more likely to be a script than one person manually downvoting hundreds of stories; adding a faked referrer URL to that kind of script would be a single LoC.

(And a site like Literotica shouldn't be voluntarily increasing the amount of data that it collects on its users. That's getting into "burn down your house to kill a spider" territory.)

And, again, this isn't about making manipulation impossible. If you're tech-savvy and motivated, NOTHING lit does will stop you. But if that's the consensus you people keep blabbing about whenever a proposal is made, you can just as well stop complaining about the problem. Because NOTHING will ever be done if NOTHING will ever be good enough.
This is about making it not as easy as it is right now. Because, right now, you don't even have to visit lit AT ALL. You can just open the Tor browser, post six lines of code into the console, and then have it do nothing but send a vote for a list of stories you specified before getting a new exit node and repeat the process. If you implement a referrer and maybe a 30 second wait-time before a cast vote counts, it would, at the very least, get rid of all the casual one-bombers.

 

[StillStunned]

The AH discussing what changes they think Lit should make reminds me of this clip:

 

[RejectReality]

Everyone assumes it's some supervillain running a script, but experience tells me it's a horde of overzealous, white-knight fans acting independently. ( Even the LW trolls are doing the same thing to some extent, but based upon a kink rather than an author ) They don't need pure trolling motivation to sign up. They'll sign up to support their darling, and the trolling is just the side effect. Every villain is the hero of their own story.

If it is a supervillain, then that's the person you're least likely to stop with roadblocks. Every control you put in place discourages the middle from bothering, and does nothing to stop the supervillain. You can spoof all the identifiers with equal ease if it's scripted. For that matter, you can automate the account generation and assign them their own identities from square one for ease of rotation and smoke-screening. Vote totals go down, and Dr. Evil's master plan gets easier/less detectable.

What we need is a return to the large sample, contest-grade sweeps that used to happen with the monthly contests. That was the bulwark that kept the trolling tamped down, and the absence of it is why the toplists have reached this point of homogeny. Most of this is simple, unsophisticated trolling, and it's not being caught because the samples being examined aren't large and diverse enough to pick it up. The contest sweeps have that extra little bit of magnification, ( Unfortunately, this probably takes out some legitimate votes as well, but eggs and omelettes ) which gets the next level of sophistication that is as far as most trolling goes.

The supervillains are usually focused on one person, not a toplist or the whole site. They don't get caught until there's a specific focus, and even then, it probably requires Laurel saying "Yeah, that's screwy, and even though I can't detect any concrete patterns, I'm wiping these 1s, 2s, and 3s from the last few months anyway." If there's somebody with a hate-on for the site using these techniques, then there's nothing anybody can do about it. It's the new normal until they get bored/die.

 

[NotWise]

Everyone assumes it's some supervillain running a script, but experience tells me it's a horde of overzealous, white-knight fans acting independently. ( Even the LW trolls are doing the same thing to some extent, but based upon a kink rather than an author ) They don't need pure trolling motivation to sign up. They'll sign up to support their darling, and the trolling is just the side effect. Every villain is the hero of their own story.

I don't think it's done by run-of-the-mill trolls. For both the Romance top list and the LS top list, you can identify October, 2024 as when it started. It's been very systematic, and since fairly early in the attacks repeated sweeps have been largely ineffective at reversing them. I can't speak to what's happened on LS, and can only know for sure what happened to my story in Romance, but the two sound very similar. Other lists have been attacked (N&N, for instance) while others have not.

The top lists are a legacy feature. They're one of the features the site has not updated, and now they look obsolete. The search facility can give readers any top list they want, though a few modifications could make it work better. I'd support a decision to simply drop the top lists in favor of the search. That could increase the server load and it would remove an easy access point for readers to find popular stories, but it would also remove a large, static target for attacks the site can't stop.

 

[EmilyMiller]

I don't think it's done by run-of-the-mill trolls. For both the Romance top list and the LS top list, you can identify October, 2024 as when it started. It's been very systematic, and since fairly early in the attacks repeated sweeps have been largely ineffective at reversing them. I can't speak to what's happened on LS, and can only know for sure what happened to my story in Romance, but the two sound very similar. Other lists have been attacked (N&N, for instance) while others have not.

The top lists are a legacy feature. They're one of the features the site has not updated, and now they look obsolete. The search facility can give readers any top list they want, though a few modifications could make it work better. I'd support a decision to simply drop the top lists in favor of the search. That could increase the server load and it would remove an easy access point for readers to find popular stories, but it would also remove a large, static target for attacks the site can't stop.

Agree with all of this

 

[RejectReality]

I don't think it's done by run-of-the-mill trolls. For both the Romance top list and the LS top list, you can identify October, 2024 as when it started. It's been very systematic, and since fairly early in the attacks repeated sweeps have been largely ineffective at reversing them. I can't speak to what's happened on LS, and can only know for sure what happened to my story in Romance, but the two sound very similar. Other lists have been attacked (N&N, for instance) while others have not.

The top lists are a legacy feature. They're one of the features the site has not updated, and now they look obsolete. The search facility can give readers any top list they want, though a few modifications could make it work better. I'd support a decision to simply drop the top lists in favor of the search. That could increase the server load and it would remove an easy access point for readers to find popular stories, but it would also remove a large, static target for attacks the site can't stop.

It's every toplist. They're all locked at the same point where the tier at 4.84 all have too many votes to take down quickly. If it was a script, they'd keep whittling down those 4.84s while simultaneously blasting the newcomers over that bar. That's not really happening. It's the hordes of white knights taking out the new competition that are higher on the list, the way it's always gone. This everyday trolling behavior has always had the same patterns, where there's a threat level, and everyone gangs up on the greatest threats, and works their way down.

Toplists updating more than once a day now means that each story gets trolled a little less than they used to, but more stories overall get hit. The target window changes more than once a day now. That's probably where the October '24 starting point everyone is pinning comes from. I'd be willing to bet that's where the once a day updates of the toplists ended in favor of 3-4 times a day.

What's missing is the regular bone-cutting sweeps that used to undo all their "good work" in the form of the monthly contest determination. Most of the consequential sweeps have been for the themed contests, and the sample is too small on those to get anything but trolling of the contest stories for the most part.

The toplists are too big of a selection point. They're not going to remove a traffic driver.