Currently we capture fanatic, voting ring or rage-type activities, by measuring the number of up-votes or down-votes between the same two users in a certain (relatively short) time period. When this is detected by the system, the votes are reversed, and I've never bothered to investigate what else happens after that - though I'm sure some form of wrist-slapping ensues.

While it is intended to be well-guarded for obvious reasons, the algorithm used is not exactly a secret, and I am sure there are plenty of examples where people are dangling (maybe intentionally, or maybe by dumb luck) just on the safe side of getting caught.

Personally I believe there are active, intentional, long-term vendettas in operation right now where users either:

  • blindly up-vote every answer from a user; or,
  • down-vote a particular user once or twice a day, regardless of topic or quality.

Now, maybe the first one is valid - if I'm operating in the same tag-sphere as some smart person, it may just be that I'm constantly enthralled with their answers, and I'm not even noticing who they're written by. (Though, even Skeet isn't perfect, and I haven't up-voted every answer I've seen him post.)

But still, can we do a better job of catching - or at least noticing - this activity over the long term in addition to the short term?

The behavior is the same, and shouldn't be tolerated, regardless of the time frame - vengeful rage serial voting, blind staker crush serial voting, and painstakingly patient serial voting are all undesirable behaviors IMHO.

I'm not saying it has to do anything automatically, as I'm well aware of the possibility of false positives. And I'm not even sure what to propose as a time frame or a vote threshold.

But if a large percentage of my up-votes or down-votes are against a particular user, and especially if on a majority of those posts I am the only one voting in that direction, I think it should at least be raised as some kind of flag to moderators.

This would work much like how other suspicious activity is highlighted (but also not dealt with automatically). If a pattern becomes obvious and persists, then they can intervene in some way (or not, but at least they were provided with the information - just like they can reject and ignore manual flags). Today, all I can do is e-mail [email protected], and unless I know exactly which person(s) are involved, there is not much they can do about it.

And you're thinking, "Meh, let some hater down-vote you once or twice a day. Unless all you're posting is utter crap, it should be statistically insignificant." You're right. But reputation is not the point, and is not the motivation behind my request. What I'd like to see curtailed is the tendency for folks to down-vote a question or answer because of who posted it, not because of what it said or didn't say. When a down-vote is personal it can leave a person second-guessing the quality of their post. Which isn't necessarily a bad thing, except when it happens repeatedly.

Sorry for the essay but I wanted to try to anticipate and cover all the objections and hole-poking.

  • So you're asking for P2P vote ratios to raise mod flags? tl;triedtoreadforcontent...
    – user7116
    Sep 17, 2012 at 13:45
  • @sixlettervariables I think so, something similar to what we do for serial down-voting within a 24-hour period, but without the automatic reversal (since, over the long term, it could be a coincidence). I just feel like more visibility into this behavior could make some of it go away.
    – Aaron Bertrand Staff
    Sep 17, 2012 at 13:46
  • I'm looking for you to bold the feature-request you're asking for rather than discussion points :)
    – user7116
    Sep 17, 2012 at 13:48
  • @six point taken, I bolded / quoted the feature request.
    – Aaron Bertrand Staff
    Sep 17, 2012 at 13:49
  • "(Though, even Skeet isn't perfect, and I haven't up-voted every answer I've seen him post)" Yeah, sure :P What exactly do you mean by trap, though? Sep 17, 2012 at 13:56
  • @BoltClock'saUnicorn I'm not sure if trap is the right word, but I think it is self-explanatory: I want this behavior to be noticed by someone and acted upon, where appropriate, rather than ignored.
    – Aaron Bertrand Staff
    Sep 17, 2012 at 13:58
  • @AaronBertrand did you experiment with this? I did and as far as I can tell, tripping serial-detection script has an impact on further votes - like lower threshold for reversal to kick in, target user rep becoming immune to votes coming from past-abuser. If that's what you're asking about then I guess it's there already
    – gnat
    Sep 17, 2012 at 14:09
  • @gnat I don't know how to "experiment" with this. The only serial-detection I have ever seen tripped is the one where rage serial down-voting occurs in a very short burst. Perhaps there are longer-term detections in place but the only one I've ever heard about is the daily one.
    – Aaron Bertrand Staff
    Sep 17, 2012 at 14:11
  • @AaronBertrand anyway, if your question is whether serial detection script has (and uses) sort of "memory" about past abuse, then the answer is it most likely has and most likely uses it
    – gnat
    Sep 17, 2012 at 14:14
  • I guess one way I could experiment with this is to serially down-vote someone's answers (one or two a day for a month) and see if it ever gets reversed. Of course if the devs know that this is already in place, they can simply close my request as status-completed. But I suspect that the behavior I'm talking about is not currently being caught.
    – Aaron Bertrand Staff
    Sep 17, 2012 at 14:16
  • I suspect that your test will be looked upon... unfavorably... by the team. But if they are okay with it, I volunteer to be your test subject. 32k rep ought to be enough for anyone, and I have enough posts that you could experiment for far longer than a month. The only downside is that the serial voting algorithm's parameters may be different on MSO.
    – Pops
    Sep 17, 2012 at 14:31
  • 13
    I've actually been secretly running this test against @PopularDemand for months; I didn't realize I was being helpful Sep 17, 2012 at 14:32
  • 1
    "Now, maybe the first one is valid " The second one too if you ignore the weasel words---which no algorithm will be able to evaluate---as there are user that post a couple (or more) crap posts day in and day out. Sep 17, 2012 at 16:01
  • 1
  • 1
    @AaronBertrand check out this investigation by Brad Larson into attempted "under the radar" downvoting.
    – Pops
    Sep 18, 2012 at 19:20

1 Answer 1


I'm a Community Manager for Stack Exchange. Let me stab at this.

But if a large percentage of my up-votes or down-votes are against a particular user, and especially if on a majority of those posts I am the only one voting in that direction[...]

Since this is one of the obvious signs of fraudulent voting, we have undisclosed means to detect this, even in the long term.

Today, all I can do is e-mail [email protected], and unless I know exactly which person(s) are involved, there is not much they can do about it.

Votes have enough self-identity to be tracked just by the victim (otherwise, how would we trace things for reputation change on both sides?). Without a victim, you have nothing to even be suspicious about, so in any scenario that provides you enough concern that something has happened, that's enough information for us to be able to investigate whether something malicious happened.

  • Thanks, I'm glad to hear that there is something in place. And I guess that's all I can really ask for, since knowing how effective it is would require knowledge I don't have (both a perpetrator and either knowledge of what is in place or that such activity has been reversed).
    – Aaron Bertrand Staff
    Sep 17, 2012 at 14:39
  • 4
    Personally I'd like to see some more automatic tools on this end. Those undisclosed means (assuming you mean what mods get, not some double secret admin tools) make it very hard to find anything but laughably obvious cases, and only after they've already been brought to your attention
    – Ben Brocka
    Sep 17, 2012 at 14:57
  • @BenBrocka, you don't need actually an automatic tools, i found many false voting cases without any tools :)
    – Lucifer
    Sep 17, 2012 at 15:05
  • @Lucifer I think Ben's suggestion is basically trying to focus the tools in a more proactive usage than a reactive one (even if it is a sort of reactive proactivity)
    – Grace Note StaffMod
    Sep 17, 2012 at 15:06
  • 3
    @GraceNote ok, I think there is one simple logical way to implement such thing, I am not whether is in current part of catching those cases or not, but you can implement a user voting-log activity, and from that log-activity, the user who is doing a only vote up to certain other users can easily(automatically) catch :)
    – Lucifer
    Sep 17, 2012 at 15:11
  • 1
    @Lucifer Something starting from the kind of stuff on the votes tab on your own user profile, perhaps?
    – Grace Note StaffMod
    Sep 17, 2012 at 20:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .