Timeline for How has RPG.se used protection? (crunching the numbers on spam and protection)
Current License: CC BY-SA 4.0
23 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 24, 2020 at 21:40 | comment | added | Rubiksmoose Mod | @Pyrotechnical hah! completely understandable and also it was a good point anyways since I shouldn't ever assume everyone will know what it means. | |
| Jul 24, 2020 at 21:38 | comment | added | Pyrotechnical | @Rubiksmoose I kind came into this thread without the just did 10 reviews mindset so I wasn't thinking of all the typical terminology associated with it. | |
| Jul 24, 2020 at 21:25 | history | edited | RubiksmooseMod | CC BY-SA 4.0 |
added 19 characters in body
|
| Jul 24, 2020 at 21:25 | comment | added | Rubiksmoose Mod | @Pyrotechnical I'm sorry that was bad form! VLQ= Very Low Quality. It's defined in the system (you can flag a post as VLQ) as "no amount of editing can salvage the post" as an example. Does that help clear things up? | |
| Jul 24, 2020 at 21:09 | comment | added | Pyrotechnical | You're using the VLQ acronynm very frequently, but I can't figure out what it means from context. Can you please define? | |
| Jul 22, 2020 at 16:12 | comment | added | Rubiksmoose Mod | @doppelgreener definitely true. I chose deleted/downvoted because it was the easiest metric to grab that I could roughly link to VLQ answers. Anything else is impossible to get data on without introducing significant subjective elements. So that is why I focused in on that. Hopefully, we could assume that most of these were unprotected after they had already left HNQ as well, so that becomes less of a variable as far as voting goes (no way to check that either though for the old posts and only roughly for the new), but that still leaves VLQ posts on the HNQ questions before protection. | |
| Jul 22, 2020 at 16:07 | comment | added | doppelgreener | Note that deleted answers weren't the only thing protection was applied in response to. It's also used when we're seeing lots of low-quality answers, which tends to come up in proximity to HNQ. Protection or removing from HNQ are both options for responding to that. In SEDE this won't necessarily show up as negatively scored answers because a question being HNQ tends to introduce lots of ambient upvotes, including for low quality answers. It might show up as many answers with disproportionately low total score—a long tail basically. | |
| Jul 22, 2020 at 5:06 | comment | added | user-781943 | I'm not so sure about that, but would be interested in reading a methodology. If it could be proven accurate then we have a great opportunity to put it into practice, and we have a hundred unprotected questions with 0 deleted answers to test it on! | |
| Jul 22, 2020 at 3:56 | comment | added | Rubiksmoose Mod | @gszavae certainly impossible to empirically predict I think, but understanding that is partially why I dug up all this data. Figuring out the best way to handle this is why we opened the other post to try to shake the tree of community wisdom and experience and see if we can come up with a good solution to get us as close as possible. | |
| Jul 22, 2020 at 3:53 | comment | added | user-781943 | It's also worth considering if, besides the question already being protected, it was possible to predict if the question would receive an answer when unprotected. There are over 100 unprotected questions with 0 deleted answers - how is the 1 that received an answer different? | |
| Jul 22, 2020 at 3:52 | comment | added | Rubiksmoose Mod | @gszavae your critiques about post age and sample size are completely valid. I do say it is only a rough analysis. I stand by that it's good enough to support the conclusion I drew from it though. | |
| Jul 22, 2020 at 3:52 | comment | added | user-781943 | Another issue is trying to compare a sample size of 21 days to a sample of nearly 4000 days. The first sample size is just way too small. In the past around 150 questions had been unprotected, and a third of them received answers of unknown quality within an unknown time period. Now we have 700 questions unprotected, and in 21 days 5 of them have received answers with unknown quality. Any comparison that doesn't involve further analysis doesn't yelid anything beneficial, at the very least we should know about the length of time and quality - in 21 days how many answers receive an answer? | |
| Jul 22, 2020 at 3:47 | comment | added | user-781943 | We should also consider if "answers" are even a good metric. For example, none of the new questions received spam answers, 3 were valid answers, and 1 was even made by a user with over 100 rep. Does this make it a good metric? It seems very misleading. Should a question that generates 5 adverts for cheap viagra from 1 rep accounts be put in the same bucket as a question that generates 5 excellent and well written replies from members with +100k rep and 5 years on the site each? I'm exaggerating, but surely you see the issue. | |
| Jul 22, 2020 at 3:43 | comment | added | user-781943 | there are a lot of statistical problems in that statement that need to be unpacked. Firstly, the oldest protected answer maybe 10 years old, but to say that those answers happened over 10 years is very misleading. What's more, 10 years ago the number of users on the site was much lower, so we SHOULD expect there to be more answers on the site now. We haven't done the analysis. For all we know the 54 answers happened 10 years ago, or within this year. For all we know 53 of those questions were protected this year. | |
| Jul 22, 2020 at 3:36 | history | edited | RubiksmooseMod | CC BY-SA 4.0 |
deleted 38 characters in body
|
| Jul 22, 2020 at 3:08 | comment | added | Rubiksmoose Mod | @gszavae That makes sense since the previous answers had time to accumulate answers over 10 years the fact that they could approach the same amount in 1/5 of that time seems to be supportive of what I'm saying: we should be careful what we protect and unprotect. | |
| Jul 22, 2020 at 3:07 | comment | added | Rubiksmoose Mod | @gszavae To me the amount of bad posts we've had from formerly protected question means that the ones that have gotten junk answers so far were correct in being protected and shouldn't have been unprotected. It doesn't mean that were all questions should have remained protected (something I've explicitly said is not true). | |
| Jul 22, 2020 at 1:51 | comment | added | user-781943 | The sample sizes are too small to make a confident decision. A third of all previously unprotected questions have received an answer after protection. At 2 per week that would take 2 and a quarter years for the 700 questions to reach the same proportion, even if we assume the last 2 weeks is representative of the next 117 weeks. | |
| Jul 22, 2020 at 1:43 | comment | added | user-781943 | For reference, out of the 5 questions that received answers the number of deleted answers by low rep users per timeframe each of them had beforehand (in brackets is the total number of deleted answers) is: 1 (2 total) in 5 years, 1 (2 total) in 3 years, 3 (3 total) in 3 years, 0 (0 total) in 3 years, 2 (2 total) in 6 years. I think we should be mindful of what metrics we are using to make smart decisions. 0.7% of the unprotected questions saw answers in 21 days. 2 out of 5 is a big portion, 2 out of 700 not so much. | |
| Jul 21, 2020 at 22:39 | history | undeleted | RubiksmooseMod | ||
| Jul 21, 2020 at 22:39 | history | edited | RubiksmooseMod | CC BY-SA 4.0 |
added 1138 characters in body
|
| Jul 21, 2020 at 22:26 | history | deleted | RubiksmooseMod | via Vote | |
| Jul 21, 2020 at 22:26 | history | answered | RubiksmooseMod | CC BY-SA 4.0 |