[ SmokeDetector | MS ] URL-only title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +1 more: mumybear.com/vitax-lean/ by phirbna12 on askubuntu.com
[ SmokeDetector | MS ] URL in title, blacklisted website in body, blacklisted website in title, pattern-matching website in body, pattern-matching website in title, +2 more: wrinklerewindfrance.com/foliactive/ by arkdler on graphicdesign.SE
the "toxic" verdict basically comes from a Google API, we can tweak the threshold but I think so far it's vaguely useful, though not particularly exciting
@tripleee That pattern looks like it's already caught by Potentially bad keyword in answer and Potentially bad keyword in body; append -force if you really want to do that.
@tripleee That pattern looks like it's already caught by Bad ns for domain in body and Bad ns for domain in answer; append -force if you really want to do that.
you never know until you know, there was this email spammer a few years ago who made the news with "I am a stranded time traveler from the future, I need the following parts to get back home ..."
It's one of our first meme, inspired by a local troll : Rondo.
This user appeared on day asking of list of movies about time traveling robots.
His post got closed pretty fast because questions calling for a list of works are off topic as per our FAQ, but he came back with variant of the same que...
I don't think we can unilaterally change to a different model, the proper way to go would be to evaluate models under multiple parameters (should multiple keywords trigger multiple rules? should body vs title be kept separate? should question v answer be kept separate? should domain v subdomain be conflated?)
@quartata I have NAAd many of the toxic flags, they are not toxic enough to deserve rude/abusive IMHO
@Undo I personally think that the first priority should be redoing the flagging condition preview so that it calculates accuracy across all posts. If a bunch of the autoflag FPs had weights like 61 then that's not a problem with that
{IMO} The problem isn't with setting low conditions. It's using a high spam site like Astronomy and adding it to a condition which includes other sites (which don't have a high tp rate) using that to circumvent accuracy.
I personally have a very low threshold for the main few spam sites, but for the rest I have a conservative number (and I don't use the high spam sites [again] in order to drop it lower)
this is also about balancing our values as an organization, "I don't worry because nothing bad happened" is fine for you as an individual but here you have us wasting time looking for ways to reduce FPs on autoflags
It's just 1/1/3 is very similar to 160/1/3, and the fact you can create put the minimum weight down to 1 allows you to do things which I think should be blocked.
@quartata ok, I decided to put data behind my words, and I've been overreacting. Since the beginning of MS, there has been 239 tps that had 3 or more reasons and a weight under 160, but only 9 fps.
> Would not be caught as a post, title or username.
> Potentially bad keyword in body, potentially bad keyword in title, potentially bad keyword in username ---------- Title - Position 1-12: inzhoop.com Body - Position 1-12: inzhoop.com Username - Position 1-12: inzhoop.com
@quartata it might be a good idea to process the scan rate and report that as a statistic to MS so we can see response times, etc. as a statistic there, rather than trace the console output
unrelated Lunar Eclipse and Supernova both have the perspective changesets, and Solar's going to get them once I do a manual pull (which I always do when I have to reboot the bot)
