Anybody have any ideas what I can do to get more flags on GD.SE....? I hit 3k and all my ~5 daily flags are now just close votes, and I still need another ~125 for Marshal. (I now get maybe 1, if I'm lucky 2, from naa) and usually none from smokey, spam quieted down there lately. (...just say I don't need it, which I don't, but I want it :S)
@GrantGarrison No, because similar to your worries about autoflags, what if a comment is added in the wrong place? And a comment is a lot more harmful than a downvote
and certain sites are exempted from certain detection reasons
but there is no whitelist within the individual detection patterns. Anywhere.
so you shouldn't ask to whitelist or exclude certain patterns from certain triggers/detections, because it's nontrivial and extremely complex to try and implement, and not likely to happen, because it will increase the code complexity by increasing magnitude beyond the already-complex code base it is that handles the reasons and detections..
and trust me I've been through the code that handles these detections, it's complex enough as it already is, to add whitelisting to each reason makes it obscenely complex.
and it's already on the verge of that level of complexity.
Sometimes, when certain keywords are blacklisted with a regex, we do add a negative lookahead/lookbehind to make sure some exceptional cases don't trigger.
Anybody have any ideas what I can do to get more flags on GD.SE....? I hit 3k and all my ~5 daily flags are now just close votes, and I still need another ~125 for Marshal. (I now get maybe 1, if I'm lucky 2, from naa) and usually none from smokey, spam quieted down there lately. (...just say I don't need it, which I don't, but I want it :S)
@WELZ Multiple flag conditions. It's possible to have multiple conditions that each pass 99.75 individually, but catch different sets of posts and end up under when taken all together.
you just find the set of posts that each of them would catch, uniquify them, and find the total accuracy
@ArtOfCode I know exactly why this doesn't work. Duplicate sites will mess with it. For example, Super User drops the weight lower, so it can't be used on it's own then used again.
@JakeSymons Your flag conditions are accurate enough, but they're kinda redundant... the 195/1000/1 condition will be matched before either of the others, unless it's on the 4 sites that the third one has that it doesn't.
and you've got two 280/1/1 conditions that differ only in site count, which is completely redundant - I'm gonna delete one of them
@WELZ Neither. If a post is caught by any of them, it's included in an array (excluding duplicates), and the accuracy of the completed array is checked
@WELZ to circumvent the accuracy limit, you can use sites such as Super User to raise the accuracy. However, when you use that more than once, it then is a duplicate and is removed in the total.
You take the set of matched posts for each condition, you merge them, and you remove duplicates. Then you count how many of the resulting set are TPs, and you divide that by the total size of the set.
The set of matched posts might be different for each condition. The union of those sets, therefore, might be lower accuracy than each set individually.
Ok so im going to use random numbers here. SU: 150 weight min. SO: 200. SO + SU: 185. Using SU+SO and SU means that you now have (I know the code doesn't do it this way) SO at 185, which it isn't at.
@Catija Yes, but the user appears to be using the "answer" solely to promote their site. The answer makes no attempt to answer the actual question asked, but immediately changes subjects onto "another common problem" in order to be able to link to their site. That's spam.
