@Undo OK, should I vote on that, or not?

yes you should

Confusing policy politics. Policics

you should use time travel to create a bunch of clones to upvote it as many times as possible

MSE is free to everyone. Not really that confusing.

Pretty good idea

@Undo metasmoke has a Smokey access token, right? Could easily automate meta post monthly reports from there

But isn't it messy for every site?

@ArtOfCode not sure it does actually

Maybe a post with monthly updates as answers?

unless it's tied to the System account

@M.A.R. nah that'd be fairly easy

@quartata pretty sure the System account has the Smokey AT for when we tested flagging

oh

for the dry runs?

I think it's time for everyone to stop commenting on Valorum's answer. If he hasn't gotten it by now I doubt he will; there's only so many times you can explain the same things over and over.

cause we did fetch the flag options right

yeah ok

I kinda like that if Shog thinks that's all right

I think it should be opt-in obviously

@quartata The meta posts?

a meta post on each site

with data from the site dashboard

I suppose Meta SE would be excluded....

Heh, Valorum's post got deleted

Thank goodness

not surprised. I suspect the mods were only somewhat tolerant of it

I wish he didn't have to resort to ad hominem, we need opposing view points

just not

you know

bad opposing view points

@quartata It was self deleted.

well then maybe he saw his mistake

i'd say that's a happier ending

@quartata not the easiness, but clogging up almost empty metas

@quartata You really don't know him, do you?

@terdon I do

I'm an optimist OK

Then I salute your optimism.

:)

@quartata nah

nice

...

Uh, well, it's probably exactly how democracy works

Forget it.

Anyway, I meant to go to sleep half an hour ago.

also the people who say "oh, SFF doesn't get much spam" I think haven't been looking at the numbers

it certainly gets a decent amount

I meant to go to sleep 2 hours and 38 minutes ago

and the deletion times are very good on it with the increased flags

there's a difference

trouble is the site meta posts aren't going to have teh graphs

@quartata You don't notice something that gets deleted due to SD.

It's easy to say "we don't get spam"... But the reality is that you don't realize it because SD already deleted it

@quartata Why not? Couldn't they have static jpgs?

no no

I... probably bug @Randal'Thor a bit by raising custom flags asking for the nuking of users after the actual spam is gone

I mean when people make posts like the one on SFF

they're not cross posting the graphs

@Undo I think the one concern that was raised that seems problematic to me is that at 5 autoflags, a post could be deleted on accident. That would not be an issue at 4

That's a valid concern, actually. Misclicks can happen.

except that FPs above 300 weight haven't really happened yet

By the way, I realized I had a mistake in the original graph: I hadn't limited the data to autoflags only (sorry, I didn't know the database structure well enough). So I made a new graph, and also made it so that the boxes represent the size of the respective datasets. Feel free to update the post or not, the main point of that graph is still the same, mostly, but this version does show a greater number of outliers which could strengthen your point:

Another thing to note is that we've been running in this mode for quite some time for 3-flag sites, with no issues

@DJMcMayhem The likelihood of the auto-flagged post being a fp to begin with is extremely low... then the like likelihood of a user misinterpreting the post and clicking spam, is again, very low... Two acts of god converging at the same time, same post, is extremely unlikely

And feel free to blame it on me if you do choose to use it and people complain about changing it in mid-discussion.

@K.Dᴀᴠɪs Fair enough. I didn't realize it was tiered either

@K.Dᴀᴠɪs and we still can flag for undeletion

@JohnDvorak not if we don't have rep?

@K.Dᴀᴠɪs and even then it's easy enough to get ahold of a mod to undelete

@quartata eh, TL, chat

involves dragging a mod in one way or another

that's not really the point we want to argue

@quartata it will be hard to catch this room in a state with no available flaggers ... while a post has just been flagged.

once it's been deleted I mean

@Mithrandir The problem is not so much undeleting but the human factor. It's a pretty bad user experience to have your good faith post deleted as spam. I wouldn't bring up ease of undeletion as an argument really. I think you have a much stronger argument in the fact that the autoflagging system works better than humans.

you would have already had to have the flag dialog open

@terdon yep yep

thanks, that was better phrased

You can flag any other post on the site

@JohnDvorak I suspect quart means you can't see it without 10k

They still have the link

And the text through MS

Can anyone explain how come the number of ttps I see on SO is so much lower? Is it because the community is so much larger that more spam is caught by regular users before smokey can even see it?

Perhaps spammers just don't dare on SO?

@JohnDvorak No, SO has the most spam of any site on the network.

it's more likely due to the fact that the websocket is fairly flakey on SO

Oh. I believe the websocket randomly drops 50% of SO posts, though

we fetch some questions from the API route to supplement

So, a thought: we've got weight & reasons & stuff that tell us whether a post is 100% definitely spam (according to all past experience, never seen a post with those criteria that wasn't); the meta posts suggests using 5 autoflags on posts at 99.99% accuracy. why not go the extra 0.01% and say 100%, specifically meaning we'll only have Smokey cast 5 autoflags when we've never seen a post like that which isn't spam? (& explaining that part as what 100% means)

Oh wow, 50? That's odd. I'd have expected SO to be the most stable.

SO doesn't have a live homepage

that's probably part of it

I think it's deliberate, to not swamp the realtime page for ... ahem ... human consumers

@doppelgreener I wouldn't be opposed to that

@JohnDvorak we should ask about that, actually

@Shog9 you around?

@quartata aye

@doppelgreener that part is carefully phrased to not lock us in to specific percentages

@quartata running for what?

@YvetteColomb SO mod

@ArtOfCode oh ok

didn't you place 3rd last election

If we want it to be 100.00%, we can (we're currently throwing around 300 weight, anyway, which is 100%), but having that there means if we find a slightly lower percentage works better for some reason, we can do that without meta complaining.

@quartata yep - are they having a new election?

yeah i wasn't sure if it was backwards-looking percentages, or forwards-looking ("there's always a chance it could be wrong but it'll be like 99.99% accurate")

mmhmm

stackoverflow.com/election/10

@quartata thanks

tenth election. Wow

gently debating whether I should run or not

you should

diplomacy is always welcome

leaning towards not under the "not enough activity"

hmm

Actually, maybe I want you focused here.

yeah don't run :P

:P

4 K seems like it'd be a tough sell on SO?

I darn near did it one year

7kish

@Catija I did not-terrible a couple elections ago with slightly less than I have now

> Human behavior is weird — like, really weird. People make everything more complicated.

accurate.

relatable

Only 12 candidates in the last one... interesting. I sort of expected more... though, I guess the higher minimum rep requirement keeps it from looking like that election on Magento.

@ArtOfCode We could. It'd take the burden off of us.

Need to figure out a schedule - maybe dynamic based on activity.

How about annual questions with monthly answers?

Or less for sites with little spam

That's more to keep track of, technically

But it'd be socially easier

I'm on vim. They basically get 0 spam and 0 meta posts

Should we do something about that? :P

Maybe have a few thresholds - 0-10 per year? Don't bother, numbers aren't useful. 10-100 works yearly, 100+ could be quarterly.

Please don't flood my site with meta posts :(

@Catija Monthly is too much, IMO.

@Undo I like that

@Undo Might be worth it on SO and/or AU. Not much outside that, though.

You wanna deal with a monthly mSO post?

Maybe one post still to make sure we're on the same page?

@DJMcMayhem mSE works for that. I'm comfortable using the current one as a proxy for the <10 volume sites.

If they had a way to search deleted posts by reason, that'd be really easy for any mod to check on their own, whenever they wanted to.

@JohnDvorak Not sure if you're proposing we flood vi.meta with numbers or vi.main with spam...

@Catija What do you mean by 'reason'?

:P

Assume you mean 'why it was deleted', not 'SD reasons'

just a note for Art: this is also something that I think falls under the duties of future Half-Life

@Undo I'm talking about SE search, first of all... So, I mean "red flag deletion"... vs self deleted or moderator deleted.

fwiw, deleted:1 locked:1 works just fine.

A few false positives in there, but usable.

separating the storage and communication and the analysis

I didn't realize that they count as "locked"...

Spam is automagically locked

To keep the mortals from undeleting it or editing it or whatever.

That's great to know... I'm mostly saying... if I don't know it... and I'm pretty damn sure that most mods don't even realize that deleted:1 works... It might help?

Yeah, it would be. We could probably put it in those meta posts in a footnote or something.

@quartata Worth considering that you'll need a way to keep tokens secure. Don't know how HL is built, but good to think about that now.

right now it isn't built like anything

hl in name only

the current hl just sends stuff down a websocket :P

MS is really really good at token storage, it's probably an okay starting place for this.

@Catija you might also want to add migrated:no to filter out migrated posts (since those also appear with deleted:1 and locked:1)

@Undo this is the context:

@terdon Good to know. We don't migrate... pretty much ever... but I'm sure other sites do (ELU, for example).

basically I'd build a dashboard for the stuff it already has

then tack on anything extra only Smokey needs

I don't want to bother generalizing stuff other bots don't need

Inferno will probably do autoflagging, but that's only because Natty needs it too

this is way far in the future though

New metasmoke user 'NH.' created

hi

[ SmokeDetector | MS ] Potentially bad keyword in answer: Recovering a Windows XP License key by Ittech Gyan on superuser.com

Hi!

sd k

Hmmm get windows key for free seems like spam to me

Copied from quora.com/Why-isnt-my-Windows-XP-product-key-working or somewhere, with the spam link added.

!!/test get windows free

> Would not be caught as a post, title or username.

>.<