[ SmokeDetector | MS ] Offensive answer detected, toxic answer detected: How to silence a noisy typer by caspianhell on workplace.SE

[ SmokeDetector | MS ] Potentially bad keyword in answer: Explanation of how list comprehension works? by Manish Methani on stackoverflow.com

fp- by ArtOfCode

tpu- by Makyen on How to silence a noisy typer [MS]

[ SmokeDetector | MS ] Bad keyword in answer: Why didn't Minato's shroud change when he used Tailed Beast Mode? by sporewarrior25 on anime.SE

fp- by K.Dᴀᴠɪs

[ SmokeDetector | MS ] Bad keyword in link text in answer, blacklisted website in answer: File path unified in Windows/MacOS/Linux by tom andreson on codereview.SE

tpu- by K.Dᴀᴠɪs

@SmokeDetector This was answered 35 mins ago and we just now see it?

Ah of course they are

@WELZ Probably should delete that so people don't get any crazy ideas

Like you?

:P

The report of SD's quota roll-over, just a few messages above this, shows that there were two posts in the queue for Code Review at that time, which would have included that post.

[ SmokeDetector | MS ] Bad keyword in answer, bad keyword with email in answer, pattern-matching email in answer: Creating duplicate features based on many-to-one conversion of related table? by chelsea on gis.SE

tpu- by K.Dᴀᴠɪs

Restart: API quota is 19584.

[ SmokeDetector | MS ] Toxic answer detected: Will increasing or decreasing wheel size speed up an object? by Orca Porca 537 on physics.SE

tpu- by K.Dᴀᴠɪs

Recovered from ssl.SSLError: [SSL: BAD_LENGTH] bad length (_ssl.c:2130)

@SmokeDetector This is a test comment

Looks OK.

[ SmokeDetector | MS ] No whitespace in answer: Player in-voice sounds in counter strike 1.6 demos by fltkg on gaming.SE

tpu- by Videonauth

[ SmokeDetector | MS ] Blacklisted username: Is there a real detection limit in NMR spectroscopy? by Anonymous Teacher on chemistry.SE

tpu- by Makyen

@SmokeDetector Typical Anonymous Teacher post. Plagiarized from here, or other listing of this article & abstract.

fp by Nisse Engström on Why didn't Minato's shroud change when he used Tailed Beast Mode? [MS]

tp by Nisse Engström

> @NotEvans-Since I have given a correct answer to your question,Please give me a bounty worth +50 reputation within 3 days from today – user65163

F*** you.

@Jamal That's a very typical statement from "Anonymous Teacher". I've been surprised the few times they haven't had something like that on one of their bounty question plagiarized answers. I certainly understand your sentiment about how it makes one feel.

Do they also commonly self-delete? I'm wondering if that's a way to avoid an IP hobble.

@Jamal At this point, yeah, they normally self-delete, probably exactly for the reason you mention.

[ SmokeDetector | MS ] Offensive answer detected: IP address in network different for LAN and Wireless by user914482 on superuser.com

fp- by Makyen

@SmokeDetector Light swearing describing a piece of equipment; repairable.

Shall we clean up unused joke commands? Like

!!/brownie

@iBug Brown!

@iBug Why is there suddenly a focus from people about cleaning up the joke commands? I expect that the response will be the same as was given in the last few days about changing the !!/blame command. Personally, I don't really care that much, but...

@Makyen Suddenly a focus from people? It should have been from iBug.

@iBug You're not a person?

@Makyen I wasn't here at the time so I don't know what you refer to by "changing the !!/blame command."

@Makyen I am a person, but not "people" (plural).

I'm afraid I'll be failing the upcoming English test...

@iBug Given that you were not the one instigating the !!/blame discussion, the use of "people" was correct. I used it to highlight that there was a discussion with another person, which had relevant information.

@Makyen Fine. You have a link to the message that I can start reading from?

That would start around here and this issue on GitHub.

[ SmokeDetector | MS ] Potentially bad keyword in body: Google Sheets URLFetchApp not defined by PokerMaster on stackoverflow.com

fp- by iBug

fp by Nisse Engström on IP address in network different for LAN and Wireless [MS]

[ SmokeDetector | MS ] Link at end of answer, linked punctuation in answer, pattern-matching website in answer: Trouble setting the WooCommerce store base page & store title by Ali Bukhari on stackoverflow.com

tpu- by iBug

[ SmokeDetector | MS ] Few unique characters in answer: Redirect subdomain to sub-folder by Simon on stackoverflow.com

tpu- by iBug

@SmokeDetector Are we over-confident to 4-flag a post with only 200 weight?

naa- by tripleee

@iBug huh? have there been 4 autoflags on something recently?

@tripleee The SD report I replied to.

Also, I suppose we've started quad-flagging a while ago.

yeah, but my understanding is that it should not trigger at 200

Yeah, me too. That's what I'm confused about.

this one had four too

4 Flags is for >99.9%. 200 weight is 99.93%, which qualifies.

do we have the limits black on white somewhere?

black on white?

Oops, one more idiom that I don't know

though I found the precise limit, there were two adjacent posts yesterday which got 194 and 195, where the former got 3 and the latter got 195

As I understand it, it's dynamically calculated at the time the flags are placed.

@iBug written down, black ink on white paper or black pixels on a white pixel background

@tripleee github.com/Charcoal-SE/metasmoke/blob/master/app/models/…

Although this is a comment, I believe relevant code are conforming it.

123234 (194) and 123235 (195)

Yes, the break-point is 195. 195 weight is 99.90% and 194 is 99.76%.

@tripleee Since you're an admin on MS, you can check the settings.

yeah, I was trying to find the place, that's why I asked (-:

BTW, can you share the values? I don't think the 5-flag threshold is 99.99% at present.

metasmoke.erwaysoftware.com/flagging/settings isn't exactly transparent

oh there

scaled_max_flags is 0,0,0,99.9,101,101

but if 195 is enough for one more flag, I'm skeptical

195 doesn't make it more vulnerable?

[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, +1 more: Keto Tone Diet Keto Tone Diet by LouPatel on askubuntu.com

tpu- by iBug

I'm thinking maybe the formula should be different -- many high-weight posts will drag the overall accuracy upwards if I understand correctly how the current ratio is calculated

I think Art and Undo have a better understanding about this

if I'm reading metasmoke.erwaysoftware.com/data/sql/queries/… correctly a limit of 259 would currently result in 100.0% by the current formula

even that sounds kind of low for 5 autoflags but I would be fine with 4 and uneasy about any limit significantly lower than that

259? I remember we have 5 FPs above 280

@tripleee No, the actual weight of a post doesn't matter, just if the weight is above the threshold being tested.

yeah metasmoke.erwaysoftware.com/data/sql/queries/… agrees

We don't do 5 flags at the moment, just 4 flags. 5 flags would require 101% accuracy (the first 101 number in the list of numbers in scaled_max_flags).

@Makyen but I mean if the precse threshold is calculated dynamically, the statistics will show a lower FP% at any particular weight over time if we accrue more TP posts with a higher weight than that

@Makyen understood, just thinking out loud about what the threshold for that ought to be if we ever take that step

@tripleee I believe there's something called homeostasis

@tripleee Yeah, that's the point of calculating it dynamically.

Unless we add more reasons that influence the weight system, we aren't in the danger of consistently increasing TP rate above a particular weight.

@tripleee The threshold is currently defined as an accuracy number, not a weight.

Based on the assumption that we get evenly-distributed characteristics in all the posts on SE>

yes, I just think the resulting weight is kind of low with the current formula

@iBug I thought I saw one but now I can't find it, the chart I linked to is the same one I linked earlier ... too early in the morning for me perhaps

Well, to be honest, my opinion is that we should be periodically re-running all posts in MS against the current detections to know what the weight distributions are for the current set of detections. We can keep the original data too (as it's of interest), but re-running to get current info is the only way to actually know where we are now wrt. what weight distributions and # detections currently are.

@iBug oh I'm looking at autoflag FPs -- so we are probably both correct

metasmoke.erwaysoftware.com/… returns 100 posts, I reviewed them all and confirmed that the max is 258

!!/report movies.stackexchange.com/a/89682/48608

[ SmokeDetector | MS ] Manually reported answer: 12 Monkeys: Season 1, Episode 3 by hasim on movies.SE

tpu- by Makyen

@ArtOfCode thanks, I'll update the Teams comment as well

However, I'd like to see those calculations use is_naa OR is_fp OR !is_tp rather than just !is_tp. Using !is_tp counts conflicted posts as not being FP, when in reality they are conflicted (i.e. unknown; & should either be completely ignored or count as FP). If you'd like to see the numbers of conflicted posts in flagging conditions, I created Flag condition TP, FP, and NAA counts.

In the ranges which we use, the difference in the calculation can be significant. For example, my flagging conditions result in an accuracy of 99.75% for is_naa OR is_fp OR !is_tp, but 99.86% for just !is_tp. In an ideal world, we wouldn't have any conflicted posts, but we actually have quite a few, which can significantly affect the numbers when we're talking differences of 0.15%.

!!/watch flixstar\.net

@tripleee Added flixstar\.net to watchlist

CI on 3dd4a33 succeeded. Message contains 'autopull', pulling...

[ SmokeDetector ] SmokeDetector started at rev 3dd4a33 (SmokeDetector: Auto watch of flixstar\.net by tripleee --autopull) (running on teward/Solar Flare)

Restart: API quota is 18677.

[ charcoal-se.github.io ] continuous-integration/travis-ci/pr failure on 51bc032: The Travis CI build failed

@Makyen Actually, to be more accurate, all I did was modify @ArtOfCode's Overall flag condition accuracy. I did so to investigate the difference in the accuracy reported for flagging conditions between when flagging conditions are disabled and what's shown on the flagging conditions creation/edit page.

@SmokeDetector That's not my fault

[ SmokeDetector | MS ] Bad NS for domain in body, blacklisted website in body, pattern-matching website in body, potentially bad keyword in body: More Together With A Weekly Supplement by hosehiko on english.SE

tpu- by iBug

[ SmokeDetector | MS ] URL in title, bad NS for domain in body, bad NS for domain in title, blacklisted website in body, blacklisted website in title, +4 more: platinumcleanserinfo.com/luminary-v/ by janitodie on english.SE

tpu- by iBug

!!/report movies.stackexchange.com/a/89682/27264

[ SmokeDetector | MS ] Manually reported answer: 12 Monkeys: Season 1, Episode 3 by hasim on movies.SE

tpu- by Makyen

!!/blacklist-keyword luminary\W?v

@tripleee Blacklisted luminary\W?v

@AJ This was also manually reported ~15 minutes ago. It still needs more flags.

[ SmokeDetector ] SmokeDetector started at rev a5b98a7 (SmokeDetector: Auto blacklist of luminary\W?v by tripleee --autopull) (running on teward/Solar Flare)

Restart: API quota is 18595.

@Makyen It's standing on 6. Perhaps there is a couple of downvotes. I think one or two flags should nuke it.

!!/unwatch luminary\W*v

@tripleee Removed luminary\W*v from watchlist

# oops

[ SmokeDetector ] SmokeDetector started at rev a5e74d7 (SmokeDetector: Auto unwatch of luminary\Wv by tripleee --autopull*) (running on teward/Solar Flare)

Restart: API quota is 18564.

[ SmokeDetector | MS ] Potentially bad keyword in body: shell programming birthday assignment by Namita Itekar on stackoverflow.com

fp- by tripleee

[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, bad keyword in title, blacklisted website in body, body starts with title and ends in URL, +1 more: Premier Diet Keto The main part by user122568 on graphicdesign.SE

tpu- by iBug

[ SmokeDetector | MS ] URL-only title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +3 more: revivagenixfacts.com/luminary-v/ by user840433 on askubuntu.com

tpu- by iBug

Since tripleee is Finnish, it should be UTC or GMT+1 there, so tripleee's staying up at 5 o'clock?

@iBug not sure what you are up to, local time here is 8 AM

Finland is GMT+3?

DST yes, normal time +2

ugh, DST again

!!/watch clickbait

@gparyani You don't have code privileges, but I've created PR#2255 for you.

Potential sockpuppetry by a suspended user on MSE. See Tavern on the Meta.

Merged SmokeDetector #2255.

CI on 4fc26c8 succeeded. Message contains 'autopull', pulling...

[ SmokeDetector ] SmokeDetector started at rev 0725a49 (metasmoke: Merge pull request #2255 from Charcoal-SE/auto-blacklist-1528952699.9887857 --autopull) (running on teward/Solar Flare)

Restart: API quota is 18431.

Can we have site-specific watches?

@gparyani have there been other posts from that user containing "clickbait?"

@gparyani Yes, manually

Also, are those posts red-flaggable? Are they worth trying to catch with Smokey, rather than leaving a custom mod flag or something?

Just before getting suspended, that user went on a spree of editing questions about Hot Network Questions with assertions they were "clickbait" and retagging a bunch of them as [clickbait]. They also proposed an offensive tag wiki for [clickbait].

SmokeDetector sees edits but not comments or tags

Would it be better if this keyword is active on Meta only?

Now we have an account that's going out and making very low-quality posts (euphemism) about clickbait. I think that's extremely suspicious.

@tripleee The recent posts by the sock puppet contained the word in the body.

@tripleee and if they've been suspended, it won't even see the edits as they'll have to go through the review queue from a sock

SD doesn't see edit suggestions, only applied edits. if you want to monitor edit suggestions, you should talk to the people in SOBotics.

@gparyani I'm not yet convinced we want to catch those -- willing to be persuaded otherwise though

@Makyen huh? we do get alerted about updated posts

@tripleee Yeah, I was in too much of a hurry and wasn't specific enough.

I think you two are mistaking "edits applied" and "pending suggestions"

oh yes, approved edits only

If this sock, or any new sock puppets they create, make posts about it, we can try to catch them. The most recent post (closed as dupe) would have been caught by it.

@gparyani but does Charcoal want to catch that?

Are those posts good for red flags?

^

I'd say yes. Repeating the same behavior that caused one to get suspended from a sock puppet is definitely abuse, and should be red-flagged IMO.

But I'd say red flags are for contents, not users. Mod flags are for users.

But red-flagging a good post that's been edited to vandalize isn't appropriate. The vandalism should be reverted and custom-flagged, if the behavior is repeated.

If it's trolling with garbage, then it's definitely good for red flags, like wingding

@Makyen Two new posts were made by the sock puppet.

@Makyen The main account is suspended so no edits can be applied directly by that troll

I'm fine with the watch if it helps curb this abuse but probably some of the SObotics bots would be better equipped for this particular task

OTOH, we are supposed to give TP feedback on edits which are considered R/A. So, we do think the edit itself should be blocked on a system level, just not red-flagged.

right now, I'm kinda thinking "we deal with it if it becomes a problem"

I don't think it's red-flaggable yet, but if we get a dozen of those I'll reconsider. Right now I think we leave this to the mods

There are only 203 occurrences of "clickbait" on SE, so it's not all that common (i.e. it won't be a flood of FP).

Watching clickbait might be worthwhile, but I don't think there's any benefit to trying to develop something new in SOBotics

I'm good with the watch, as long as it's removed when this flood of trolling is over

I agree.

This guy has not yet demonstrated an ability or desire to evade bans (besides creating a single potential sock), he's not on a level anywhere near our favorite seasonal acquaintance

I think watching clickbait is meaningful as it helps us keep an eye, but I don't think we should even be red-flagging yet, this is something mods can deal with

TBH, I don't even know what this guy is.

:45149208 not quite spelled that way, but yes :)

The 2014 hit by Calvin Harris?

When did I become a RO?

probably not as important as it was a month and a half ago, but likely best to stay away from it

[ SmokeDetector | MS ] Bad NS for domain in body, bad NS for domain in title, link at end of body, pattern-matching website in body, pattern-matching website in title: check outwww.health4zone.com/hyperion-male-no2-booster/ by wsadozjerry on superuser.com

tpu- by iBug

oh

congrats :)

Some pointers to the potential sockpuppeteer: they're a former SO moderator who resigned as a mod just before taking employment at SE, then laid off as part of the November 2017 layoffs but retaining their moderator privileges on SO, and subsequently resigned as an SO mod again.

ouch, this could get ugly

@gparyani ...you're kidding

resigned in the wake of welcoming?

@NobodyNada Yes

...no way

There's a user who keeps a log of SO moderators, you can see who I'm talking about in that chart

:(

@gparyani I know who you're talking about

This person definitely used to be sane

Aren't all SO mods recognised by a "Sheriff" badge?

[ charcoal-se.github.io ] continuous-integration/travis-ci/pr failure on 3577ea2: The Travis CI build failed

wait, what? A previous SO mod has gone trolling?

Yes

I...can't believe this

Their MSE account is suspended.

He used to be super-helpful and involved here

@NobodyNada I can't, either. That's ASTONISHING.

He's still being productive and helpful on both SO and Meta.SO

It reminds me of Yvette, who used to be suspended year-long and is now an SO mod...

[ SmokeDetector | MS ] Mostly non-Latin answer: sharedfilelistd and distnoted hog CPU by user1707683 on apple.SE

naa- by iBug

@iBug It was a 20 year suspension, actually.

@Makyen Why?

@SmokeDetector Typical "Me Too" comment posted as an answer, so definitely NAA.

@gparyani I wasn't around for it, so I don't know the details. All I know is from reading some associated material. IIRC, there was a meta.

IIRC Yvette briefly touched upon it in her presentation when she ran for mod, but basically, water under the bridge

Can you do SEDE queries now? I can't log in, nor can I do CAPTCHA.

there was a bug related to cookies, try clearing those

There's no cookie. I've re-opened SEDE in an incognito window

I'm able to log in

and run queries

Log in with what?

FWW meta.stackexchange.com/questions/309179/…

@iBug the usual StackExchange login, OpenID I believe

No SE OpenID option there. I only see 6 alternative OpenID options.

[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, blacklisted website in body, link at end of body, pattern-matching website in body, +2 more: New Light CBD bone growth hip pain by ruthbrown on meta.SE

tpu- by iBug

if I click "log in using Stack Exchange" it briefly displays the SE OpenID URL and then goes to the Stack email/password page

@gparyani interestingly, the suspected sock has had accounts on other sites for about a year

No SE option, sadly.

@iBug any JS errors?

Uncaught ReferenceError: $ is not defined

I guess jQuery failed to load?

I guess, no other errors?

A lot of others, but all appear to be jQuery-related

jQuery should be loaded from ajax.googleapis.com/ajax/libs/jquery/1.7.1/jquery.min.js

Yes. I'm working on it.

@iBug check the jQuery request on the network tab?

Done. Thanks

👍

anyway, I'm off to bed now.

Good night, N

@NobodyNada It was probably created as a legitimate alt account for testing, then used as a sock puppet here

I still can't figure out who it is by looking at SEDE

Someone who types with their own human hands.

I honestly don't know where the chart is

@iBug Linked at normalhuman.github.io

oh, not SE'f chart