![]() | Essays Low‑impact ![]() | |||||||||
|
|
|||
This page has archives. Sections older than 28 days may be automatically archived by Lowercase sigmabot III when more than 3 sections are present. |
Content creation and writing experience is a reasonable requirement at RFA
The hostility of certain 'crats to some voters' preference for candidates with meaningful content creation and writing experience is bizarre. There's a current RFA that is essentially the opposite of Liz -- lots of content creation, little admin experience -- and the candidate is sailing through. It's time for the 'crats to acknowledge that many volunteers in this community value content creation as a method for gaining the particular kind of trust that is appropriate to acquire the admin tools.
Support
- (as proposer) Townlake (talk) 15:19, 7 August 2015 (UTC)
- Not an endorsement of the expectation itself, but of the fact that it is not inherently an unreasonable one. Opabinia regalis (talk) 18:39, 7 August 2015 (UTC)
- People who want to police content with admin tools should at least know how to produce it, and the only way they can demonstrate they know is to have done so in at least sufficient quantity for it to be measurable. Kudpung กุดผึ้ง (talk) 05:45, 8 August 2015 (UTC)
- I don't think anyone is suggesting that admins don't need to have produced enough content to show that they can do it, but recent RfAs have been moving in the direction of multiple oppose votes for those who have created content but who primarily do wikignome work, And in the case of Liz there were multiple oppose comments based on nothing more than her expressed opinion that wikignomes and content creators should be held to the same standard. And now I see 'crats who share that opinion being called hostile and their opinions called bizarre. That is not a neutrally-worded question. --Guy Macon (talk) 07:23, 8 August 2015 (UTC)
- With Opabinia's caveat. I'd like such !votes to clearly explain why they view the content creation as important to the RfA process (e.g "This candidate's record at making FAs suggests they have a good understanding of our article policies and work well with other people", not just "Has lots of FAs") - many such !votes miss that. Also, concerns about misusing admin tools or any "core" aspects should carry more weight than experience concerns which what such !votes (when filled as opposes) usually are. Jo-Jo Eumerus (talk, contributions) 11:35, 8 August 2015 (UTC)
- Specifically not as a "badge" that "they write," but as a sign that they are likely aware of the policies and standards appropriate to that area. IOW, we have no basis to assume that a person who has edited on few or no articles subject to WP:BLP would have the slightest concept of the strictures of that policy. A person who has never edited in any remotely controversial area might not necessarily be aware of the backgrounds of some of the more "noteworthy areas" and not have a full understanding of the policies and standards appropriate thereto (including WP:NPOV and WP:RS). A person who routinely echoes "delete" at AfD might not understand that many such articles should be retained, or a person who rarely !votes "delete" might have the exact same problem in reverse. The broader the person's evinced understanding of policies and standards,the better. Collect (talk) 12:56, 9 August 2015 (UTC)
- @Tryptofish - I did not read this as supporting any definite numbers at all, only that sufficient edits demonstrating an understanding of the main policies and guidelines is pretty much a reasonable requirement. A person who corrects 10,000 spelling errors may not have the foggiest idea what NPOV, RS, BLP etc. even mean. Conversely, a person who runs the gantlet for a single well-written article (not even GA should be "required") who has had to deal with the main policies, should not be debarred. Collect (talk) 18:50, 13 August 2015 (UTC)
- I agree with you entirely about that, and I actually like what you just said better than the wording of the proposal at the top of this section. I made my comment below partly out of discomfort about characterizing the Crat Chat as having been "hostile" to content creation-based rationales, because I think that what the Crats actually said was considerably more nuanced than that. --Tryptofish (talk) 20:04, 13 August 2015 (UTC)
- @Tryptofish - I did not read this as supporting any definite numbers at all, only that sufficient edits demonstrating an understanding of the main policies and guidelines is pretty much a reasonable requirement. A person who corrects 10,000 spelling errors may not have the foggiest idea what NPOV, RS, BLP etc. even mean. Conversely, a person who runs the gantlet for a single well-written article (not even GA should be "required") who has had to deal with the main policies, should not be debarred. Collect (talk) 18:50, 13 August 2015 (UTC)
- Do I really need to put all of my reasoning down? It's fairly clear how I feel. GregJackP Boomer! 19:19, 25 August 2015 (UTC)
Oppose "Comment"
- This begs the question and codifies what some people see as a reasonable requirement supported by consensus and others (including some 'crats -- and me) feel is a stupid, stupid requirement that has almost nothing to do with what the actual tasks that admins are asked to do. I find the wording "The hostility of certain 'crats to some voters' preference for candidates with meaningful content creation and writing experience is bizarre" to be offensive. Why not write "The hostility of certain editors to some voters' preference for treating wikignomes with the same respect as content creators is bizarre" if you are going to write a question that takes sides and insults the opposition? I have to admit, this one pisses me off and I doubt that I am alone in that view. --Guy Macon (talk) 01:42, 8 August 2015 (UTC)
- Just commenting...I do create, and I value content creators. But I also value those who perform other tasks. Some people do not enjoy writing. Some are not good at it. That doesn't mean they aren't good at any number of things that would make a good admin. valereee (talk) 20:28, 8 August 2015 (UTC)
- There is one factor that is often overlooked when editors assess a candidate's skill in content: discussion. Anyone with a few months of experience on Wikipedia can play Google News, Books, and Scholar Jenga easily, but to actually discuss content requires much more experience. Esquivalience t 23:14, 8 August 2015 (UTC)
- I'm not sure whether this is the 4th comment or the 4th oppose, but please take it as a comment. Of course it's reasonable to value content work; opposing that seems like opposing apple pie. But I don't want to argue that it's appropriate to set a specific criterion about just how much is enough. Someone can demonstrate ample skills on article talk pages without having gotten a specific number of GAs. --Tryptofish (talk) 23:46, 11 August 2015 (UTC)
- I support this partly. A 'reasonable expectation' absolutely, but I think stating that it's a "requirement" goes too far for me. It might also be useful to specify what we mean by "content creation" (from nothing to GA, from stub to GA, from stub/nothing to B class...). Callanecc (talk • contribs • logs) 03:06, 11 August 2015 (UTC)
- Speaking as someone whose original admin background was in content improvement rather than content creation, I think this is a red herring, and that while some people think it *should* be a requirement, there are many cases where it's completely meaningless. There is definitely a need for administrators who have different experiences rather than a focus on content. Examples would include someone who is particularly deft at template creation and modification and who is interested in improving MediaWiki interfaces (which really *could* use a lot of work) and management of high risk templates; and I think we really could use a quite a few more admins who are particularly focused on the identification and management of sockpuppets and vandals. We need admins who are particularly skilled at interpersonal communication, as well. We need to be giving the admin tools to people who, a year from now, will probably make good checkusers and oversighters - we've pretty much cleared out that particular talent pool, and as experienced CUs and OSs seek to move to other areas or become far less active, we don't have a lot of people in the wings who are going to be genuinely good candidates to step up and fill those roles. They have to come from the administrator pool because of restrictions on who can gain access to those tools (must have passed an RFA or RFA-equivalent process) and the need for them to have admin tools in order to do the jobs properly (deleting and blocking, in particular). Risker (talk) 20:29, 13 August 2015 (UTC)
- I think that's a very good point: that content improvement can be every bit as valid, if not more so, as content creation. Increasingly, it is becoming less important to start a page than to improve an existing page, and either kind of experience can demonstrate that an editor knows how to navigate content disagreements. --Tryptofish (talk) 22:14, 13 August 2015 (UTC)
- Content work is important, and I'm sure everyone at RFA values it and respects it in a candidate, but there is a wide disagreement as to the minimum content requirement for a new admin. I'm one of those who thinks that an admin should have an understanding of reliable sourcing, both the technical skill of knowing how to do an inline cite and the more academic one of knowing the difference between a reliable and an unreliable source. But on at least one recent RFA that put me in the support column when most of the opposes were for the candidate's low level of content contribution. ϢereSpielChequers 19:09, 14 August 2015 (UTC)
WP:NOBIGDEAL should immediately be marked historical
There is widespread disagreement over what NOBIGDEAL means. None of the interpretations are necessarily wrong, but those interpretations should be spelled out by the people using them instead of using the NOBIGDEAL shorthand that may be off-point depending on who you ask.
Support
- (as proposer) Townlake (talk) 15:19, 7 August 2015 (UTC)
- It may have been true at one time, but it is a lie now. GregJackP Boomer! 19:22, 25 August 2015 (UTC)
Comment: I think it's "NOBIGDEAL" to an experienced editor who's been around a long time (see: Wikipedia:Requests for adminship/Cyclonebiskit 2). But I think Adminship, in general, is a "big deal" – it's a job with a lot parts, and a lot of responsibility. That has only gotten more true over time. So I'm ambivalent about this – it shouldn't be a big deal, but it is, and has only gotten more so over time due to a variety of factors (including some mistakes made by the Admin class themselves). --IJBall (contribs • talk) 15:44, 7 August 2015 (UTC)
- Somebody didn't read the rules for posting on this page. Please review them, then delete both of these comments. Townlake (talk) 15:47, 7 August 2015 (UTC)
- I see that we shouldn't oppose, Townlake, so I'll pull my oppose from your first one. But I don't see anything about not commenting about refinements. If that's the case, I'm sure Dweller will be along to set me straight. --IJBall (contribs • talk) 16:01, 7 August 2015 (UTC)
- The preferences of the person who posted the survey are not binding upon those who respond -- see WP:OWNERSHIP -- and calling an oppose opinion a comment is simply playing games. If someone wants me to only express support to a series of questions that, in one case, I find deeply offensive and biased, then they should post it in their userspace where they can assert control of the responses. --Guy Macon (talk) 01:51, 8 August 2015 (UTC)
- Come on Guy, you set up ground rules at your RfCs about religion in infoboxes, and they were quite well followed. This survey was set up with the express intention to gather input on which points could be debated and which ones not. Not supporting (by not writing anything) is enough to let the proposal die. The creator of this page asked people not to oppose anything at this stage. Just that. (To avoid cluttering the whole thing up beyond recognition, as is underway.) I'm disppointed that even experienced users failed to understand the purpose of this experiment. I'm especially disappointed by the behavior of you and all the admins among the !voters. The deliberate derailing of any proposal of debate is something I see here for the first time in my almost nine years of activity. Sad, sad, sad. Kraxler (talk) 02:03, 8 August 2015 (UTC)
- I didn't try to tell people that they couldn't disagree. Nonetheless I changed my Oppose to a "Comment". --Guy Macon (talk) 07:09, 8 August 2015 (UTC)
- Come on Guy, you set up ground rules at your RfCs about religion in infoboxes, and they were quite well followed. This survey was set up with the express intention to gather input on which points could be debated and which ones not. Not supporting (by not writing anything) is enough to let the proposal die. The creator of this page asked people not to oppose anything at this stage. Just that. (To avoid cluttering the whole thing up beyond recognition, as is underway.) I'm disppointed that even experienced users failed to understand the purpose of this experiment. I'm especially disappointed by the behavior of you and all the admins among the !voters. The deliberate derailing of any proposal of debate is something I see here for the first time in my almost nine years of activity. Sad, sad, sad. Kraxler (talk) 02:03, 8 August 2015 (UTC)
- The preferences of the person who posted the survey are not binding upon those who respond -- see WP:OWNERSHIP -- and calling an oppose opinion a comment is simply playing games. If someone wants me to only express support to a series of questions that, in one case, I find deeply offensive and biased, then they should post it in their userspace where they can assert control of the responses. --Guy Macon (talk) 01:51, 8 August 2015 (UTC)
- I see that we shouldn't oppose, Townlake, so I'll pull my oppose from your first one. But I don't see anything about not commenting about refinements. If that's the case, I'm sure Dweller will be along to set me straight. --IJBall (contribs • talk) 16:01, 7 August 2015 (UTC)
- I've quoted Newyorkbrad a time or two (although I don't have the link in front of me) when he called it "a medium deal" and I agree, but "no big deal" is a philosophy that is shared by too many people to force the label of historical on. Dennis Brown - 2¢ 13:38, 13 August 2015 (UTC)
Recent RFA
The bigger issue is that any review of the data in this RfA would conclude that consensus has not been established, There continues to be a trend of change. In this case, a crat declared canvassing in the last day but the data doesn't support any change. Linear correlation is very high. A 3 day extension would clearly show whether the trend continued, was affected by canvassing or whether consensus really fell below minimum levels. What is clear is that a hard close with -3.75% per day slope at a 99% confidence in that correlation is not "consensus" of a pass. In fact, it's the antithesis of consensus and the RfA should have been left open until it was clear. Crats should use these stats to decide whether a) canvassing is significant (it wasn't here) and b) whether the approval/disapproval has flattened out (it didn't here). Any competent person would conclude that his RfA was not successful and the call for onsensus should have been pushed out. --DHeyward (talk) 05:52, 19 August 2015 (UTC)
- Consensus is not just a matter of counting votes, even weighted by people marking their !vote as weak. It's also about the issues raised and whether they are refuted or not. As for how significant the canvassing was, there were 281 !voters in that RFA, a hundred more than any successful RFA in the last 18 months, and we know that around the time the support percentage began to slip it was publicised on both a pro Gamergate site and an Anti Wikipedia one. I'm not aware of any theory other than the canvassing/off wiki publicity that would plausibly explain the very high turnout. If you look at the other eleven RFAs (out of around two thousand) that made Wikipedia:Times_that_200_Wikipedians_supported_an_RFX they were mostly when the community was larger or the candidate was higher profile or both. ϢereSpielChequers 07:50, 19 August 2015 (UTC)
- The trend line would suggest canvassing was not significant. The data doesn't reach a conclusion about pass/fail, rather the very strong linear correlation of support percentage suggests it was not ready to close. Anything external such as canvassing would alter that trend and it did not.. At some point, the R^2 correlation will cease to be that strong. I propose that mathematically, consensus cannot be determined until there is a significant shift. No shift means th consensus is still forming. A shift can indicate canvassing or consensus but it's incorrect to claim such a thing when the correlation is so high. --DHeyward (talk) 09:17, 19 August 2015 (UTC)
- Clearly the majority of the crats who participated read that RFA differently to you. I supported, and broadly agree with their close. What I see is an RFA which ran in three phases, phase one, prior I understand to the off site postings, the RFA was heading for a very clear result, there was some opposition for the candidates lack of FAs or GAs but as normal for otherwise qualified candidates whose content contributions don't include FAs or GAs, it was heading for a clear success. Phase 2 after the offsite postings, for about four days I don't recall any new significant arguments emerging, but an unusually large number of !votes came in and the support level fell sharply. Phase three, the final day, there was an RFA heading to a close in the discretionary area, so I along with many others came along, assessed the candidate, and most of us supported. Without the offsite postings it is unlikely that the RFA would ever have dropped as low as 80%, the pattern set in the first 48 hours would have continued for the next five days, the total participation would have been much more normal, and I for one wouldn't have !voted. ϢereSpielChequers 10:36, 19 August 2015 (UTC)
- The idea that there were "phases" in participation that affected voting patterns isn't borne out by the data. Such phases, if they existed, would be observable in the overall support percentage plot as sections of the line with noticeably different slopes from each other. Take another look at the graph. They aren't there. -Starke Hathaway (talk) 14:32, 19 August 2015 (UTC)
- There are three lines on that graph, and of those three the support line shows the three phases most clearly. The Oppose line at first glance seems to show a very different picture, steadily rising throughout the 7 days, but that needs to be seen in context of normal RFAs such as this one, normally after the first 48 hours the RFA calms down, most of the voting has happened, and unless new things arise the support/Oppose ratio doesn't change. Apart from the canvassing and the move into the discretionary zone there were no major new issues that came up in that RFA - Opposes at the end were still for lack of content contributions, including opposes such as 54 which wasn't struck, but whose argument was refuted. ϢereSpielChequers 17:59, 19 August 2015 (UTC)
- Yes, there are three lines on that graph, but since the one I and DHeyward are explicitly talking about is a function of the other two (and arguably the most important in determining whether a consensus had formed and stabilized), I am not sure why you seem to think "three lines" is a refutation of our argument. Yes, support votes accrued somewhat more quickly at the beginning and end of the RFA, but in neither case did those small blips (which could well have been random statistical fluctuations) significantly disturb the overall support percentage trend, which was uniformly negative over the entire period of the RFA, even the first 48 hours. No one is arguing that this was a typical RFA. The argument is that the RFA closed before consensus was formed, which would have manifested in the above plot as the slope of the overall support percentage plot line approaching zero. That this equilibrium was not reached is obviously true on the basis of the data presented. You are entitled to your own opinion but not, as they say, your own facts. -Starke Hathaway (talk) 21:18, 19 August 2015 (UTC)
- There are three lines on that graph, and of those three the support line shows the three phases most clearly. The Oppose line at first glance seems to show a very different picture, steadily rising throughout the 7 days, but that needs to be seen in context of normal RFAs such as this one, normally after the first 48 hours the RFA calms down, most of the voting has happened, and unless new things arise the support/Oppose ratio doesn't change. Apart from the canvassing and the move into the discretionary zone there were no major new issues that came up in that RFA - Opposes at the end were still for lack of content contributions, including opposes such as 54 which wasn't struck, but whose argument was refuted. ϢereSpielChequers 17:59, 19 August 2015 (UTC)
- The idea that there were "phases" in participation that affected voting patterns isn't borne out by the data. Such phases, if they existed, would be observable in the overall support percentage plot as sections of the line with noticeably different slopes from each other. Take another look at the graph. They aren't there. -Starke Hathaway (talk) 14:32, 19 August 2015 (UTC)
- Clearly the majority of the crats who participated read that RFA differently to you. I supported, and broadly agree with their close. What I see is an RFA which ran in three phases, phase one, prior I understand to the off site postings, the RFA was heading for a very clear result, there was some opposition for the candidates lack of FAs or GAs but as normal for otherwise qualified candidates whose content contributions don't include FAs or GAs, it was heading for a clear success. Phase 2 after the offsite postings, for about four days I don't recall any new significant arguments emerging, but an unusually large number of !votes came in and the support level fell sharply. Phase three, the final day, there was an RFA heading to a close in the discretionary area, so I along with many others came along, assessed the candidate, and most of us supported. Without the offsite postings it is unlikely that the RFA would ever have dropped as low as 80%, the pattern set in the first 48 hours would have continued for the next five days, the total participation would have been much more normal, and I for one wouldn't have !voted. ϢereSpielChequers 10:36, 19 August 2015 (UTC)
- The trend line would suggest canvassing was not significant. The data doesn't reach a conclusion about pass/fail, rather the very strong linear correlation of support percentage suggests it was not ready to close. Anything external such as canvassing would alter that trend and it did not.. At some point, the R^2 correlation will cease to be that strong. I propose that mathematically, consensus cannot be determined until there is a significant shift. No shift means th consensus is still forming. A shift can indicate canvassing or consensus but it's incorrect to claim such a thing when the correlation is so high. --DHeyward (talk) 09:17, 19 August 2015 (UTC)
DHeyward, you were rehashing Liz's RfA on Jimbo's talk page yesterday and she asked you to stop, so why are you doing it again here? Without looking at the voting patterns of any other RfAs your conclusions are meaningless in any case. Opabinia regalis (talk) 19:38, 19 August 2015 (UTC)
- The criticism he is presenting is about the crats' actions, as to whether there was actual consensus in any meaningful way, not anything about whether Liz's qualifications were sufficient in that RFA. Besides, discussion decisions aren't ruled by others' feelings alone, like I wouldn't ask you to retract your statement here because DHeyward had previously told you to stop bringing it up. We are all mature adults here. 108.52.24.214 (talk) 20:05, 19 August 2015 (UTC)
- The discussion at this page was started as a follow-up to the Crat Chat for that RfA. It seems to me to be appropriate to discuss what can be learned from the experience of that RfA, in terms of how better to assess the community's views in future RfAs, but it seems inappropriate to use this discussion to complain that the Crats got it "wrong". Unfortunately, most of what I read above seems closer to the latter. --Tryptofish (talk) 21:28, 19 August 2015 (UTC)
- We have a time limit on RfXs. If we were to keep them open indefinitely, I would expect a convergence to 50%, as Under the assumption that most of the people who know the candidate respond early, and later entrants are people who know the candidate less, and more likely to be uniformly distributed. An oversimplification, but the point stands. -- Avi (talk) 21:49, 19 August 2015 (UTC)
- I'm not sure why percentage should converge at 50%, even over infinite time, unless you start with the assumption that the votes are random. But regardless, no one is proposing that RfAs be held open for infinite time. The argument is that an RfA that demonstrates a uniform, uninterrupted downward trend in support percentage (that explains over 98% of the variance in support percentage!) has not yet reached consensus and therefore should not yet be closed. -Starke Hathaway (talk) 22:03, 19 August 2015 (UTC)
- @Avraham: that would only be true if support/oppose were random. That is not the case nor the presumption for consensus. A non-controversial consensus is 80%. Laws of large numbers should prevent the tails you are concerned about. Participation will tail off as qualified particpants dwindle. See the Cyberpower RfA I left on your talk page. --DHeyward (talk) 02:05, 20 August 2015 (UTC)
- @DHeyward:Let me expand on what I posted in the small text. The hypothesis is that for any candidate, there are a certain number of people who know the candidate, and many more who do not. The former group is not necessarily evenly distributed; almost certainly not. However, the latter group—those who do not know he candidate, it is more likely that there are an even number of such people. If that is representative, then I believe the distribution will tend to 50%. The N who know the candidate are broken into their oppose and support groups which may be vastly different, and they will be the ones who tend to respond early, causing the initial distribution. But the X who do not know the candidate will tend to trickle in over time, and they will do so more evenly. It is also likely that X >> N, as Wikipedia is just to big for us to all know each other any more. Once again, if the hypothesis is representative (it's certainly wrong per George Box :) ), then the initial distribution will tend to 50% over time, as those who know the candidate and have already voted are now constants, and the numerator and denominator of support and oppose continue to increase at the same rates due to the influx of newcomers who do not know the candidate as well. Of course, that is all based on the described hypothesis. An alternative hypothesis would be that the majority who don't know the candidate will exhibit herd mentality and tend to cluster to the the larger of support or oppose. My personal hunch? Cases where supports outnumber opposes by 8 or 9 to 1 will exhibit convergence to 1; cases where supports outnumber opposes by between 2:1 and 8:1 (or so) will exhibit convergence to 50%, and cases where supports outnumber opposes by less than 2:1 will fall below 50% (or be closed as SNOW before it gets there). This is not based on any data, though. It might be an interesting project for a statistician who wants to write a paper on Wikipedia, no? (paging Opabinia regalis
). -- Avi (talk) 04:34, 20 August 2015 (UTC)
- Ha. Based on recent evidence I think we can predict PLOS ONE will publish it! :) Opabinia regalis (talk) 16:44, 20 August 2015 (UTC)
- @DHeyward:Let me expand on what I posted in the small text. The hypothesis is that for any candidate, there are a certain number of people who know the candidate, and many more who do not. The former group is not necessarily evenly distributed; almost certainly not. However, the latter group—those who do not know he candidate, it is more likely that there are an even number of such people. If that is representative, then I believe the distribution will tend to 50%. The N who know the candidate are broken into their oppose and support groups which may be vastly different, and they will be the ones who tend to respond early, causing the initial distribution. But the X who do not know the candidate will tend to trickle in over time, and they will do so more evenly. It is also likely that X >> N, as Wikipedia is just to big for us to all know each other any more. Once again, if the hypothesis is representative (it's certainly wrong per George Box :) ), then the initial distribution will tend to 50% over time, as those who know the candidate and have already voted are now constants, and the numerator and denominator of support and oppose continue to increase at the same rates due to the influx of newcomers who do not know the candidate as well. Of course, that is all based on the described hypothesis. An alternative hypothesis would be that the majority who don't know the candidate will exhibit herd mentality and tend to cluster to the the larger of support or oppose. My personal hunch? Cases where supports outnumber opposes by 8 or 9 to 1 will exhibit convergence to 1; cases where supports outnumber opposes by between 2:1 and 8:1 (or so) will exhibit convergence to 50%, and cases where supports outnumber opposes by less than 2:1 will fall below 50% (or be closed as SNOW before it gets there). This is not based on any data, though. It might be an interesting project for a statistician who wants to write a paper on Wikipedia, no? (paging Opabinia regalis
- No. You can say nothing of the sort without comparing to a larger dataset. All of this stuff about lines and slopes and phases and correlations is just tasseography if you can't be bothered to use a control group. And, well, yeah, oppose votes are pretty good predictors of support percentage, aren't they? One of two key variables, I'd say. Opabinia regalis (talk) 22:37, 19 August 2015 (UTC)
- I only bring up the r square figure to indicate that the model fits the data. No one is trying to predict support percentage from number of oppose votes. They are pointing out that the support percentage trends uniformly downward over time with a high degree of confidence. This is an unequivocal indication that support percentage had not reached an equilibrium. The argument by extension is that consensus has not been established. A control group is not required to demonstrate a trend in a time series data set. Comparisons to reading tea leaves only indicate that you don't understand the argument. -Starke Hathaway (talk) 22:47, 19 August 2015 (UTC)
As the inventor of both statistics and lying on the internet, I think you'll think my credentials are better than yours.Striked out per Starke Hathaway's removal of his credential boasting. PeterTheFourth (talk) 22:54, 19 August 2015 (UTC)- I'm not trying to speak for Opabinia regalis, but I think one point is that this graph represents one incident. We would need to know if contested RfAs ever come to an equilibrium or a consensus eventually. It would be more useful to chart all recent RfAs that weren't snow cases and see how often an equilibrium was reached. As for someone who went through the experience, if an RfA ran 10 days or two weeks, I'd never have done it. Those were the longest seven days of my life (plus the two days of the 'crat chat). At some point, the bureaucrats need to stop and assess the votes and comments and right now, that time period is approximately 7 days. Liz Read! Talk! 23:09, 19 August 2015 (UTC)
- (edit conflict) Yep, Liz is right. That's not the point. Yes, it's linear; what you haven't done is demonstrate why that matters. Go look at more RfAs and see if a linear downward trend is a common behavior. You haven't shown that "support percentage equilibrium" means consensus in the Wikipedia-term-of-art sense. Opabinia regalis (talk) 23:12, 19 August 2015 (UTC)
- I only bring up the r square figure to indicate that the model fits the data. No one is trying to predict support percentage from number of oppose votes. They are pointing out that the support percentage trends uniformly downward over time with a high degree of confidence. This is an unequivocal indication that support percentage had not reached an equilibrium. The argument by extension is that consensus has not been established. A control group is not required to demonstrate a trend in a time series data set. Comparisons to reading tea leaves only indicate that you don't understand the argument. -Starke Hathaway (talk) 22:47, 19 August 2015 (UTC)
- We have a time limit on RfXs. If we were to keep them open indefinitely, I would expect a convergence to 50%, as Under the assumption that most of the people who know the candidate respond early, and later entrants are people who know the candidate less, and more likely to be uniformly distributed. An oversimplification, but the point stands. -- Avi (talk) 21:49, 19 August 2015 (UTC)
- The discussion at this page was started as a follow-up to the Crat Chat for that RfA. It seems to me to be appropriate to discuss what can be learned from the experience of that RfA, in terms of how better to assess the community's views in future RfAs, but it seems inappropriate to use this discussion to complain that the Crats got it "wrong". Unfortunately, most of what I read above seems closer to the latter. --Tryptofish (talk) 21:28, 19 August 2015 (UTC)
Is the aim of this discussion to in some way persuade the bureaucrats to reverse their decision, or to establish a principle to be upheld in future RfAs? PeterTheFourth (talk) 22:28, 19 August 2015 (UTC)
- I have absolutely no idea why people want to talk so much about this RfA, PeterTheFourth. I'm not offended, I just find it baffling that other editors have any interest at all in my editing history and patterns of editing. It's obviously become a discussion about more than me and become a conflict about who should become admins on Wikipedia. After all of these thousands of words on this, adminship clearly is "a big deal" to many people. Liz Read! Talk! 23:14, 19 August 2015 (UTC)
This isn't meant to single out one RFA. I did this RfA as well and published.
![](https://upload.wikimedia.org/wikipedia/en/thumb/7/75/CyberPower678_RfA.jpg/220px-CyberPower678_RfA.jpg)
It's point is to highlight the need for tools to help establish when consensus is achieved and when it is fluid. CyberPower678's RfA was not changing much. The RfA was very much in the discretionary zone and the data doesn't indicate it was changing much. My perception is that this RfA was ripe for closing and well within discretion. The percentage was unlikely to change significantly in any predictable way. Tools like this that would help crats determine when a level of consensus is reached and an RfA has reached the end of its usefuleness is important. This isn't second guessing the result, it's analyzing participation for consensus. A "vote" would cut off a 99% R^2 linear correlation. Consensus would extend the RfA until that went away. Cyberpower is what a non-correlating RfA where consensus is obtained. That percentage is in the discreionary zone. Simply put, Cyberpower was ripe for closing while Liz was not. No judgement is made on the judgement, only the decision to close. Both RfA's had approximately the same level of support. --DHeyward (talk) 02:05, 20 August 2015 (UTC)
- If you look at Wikipedia:Successful_requests_for_adminship you'll see that
CyclonebiscuitCyberPower678 had a pretty normal number of support votes for a successful RFA in 2014/2015and a Support percentage well above the discretionary zone. Liz's RFA had the highest number of supports this year and the twelfth highest support !vote in more than two thousand RFAs, but a support percentage that was clearly in the discretionary zone. I would not describe those two RFAs as having approximately the same level of support byeither by percentage orabsolute number. More broadly consensus is not just a vote, it is or should be a discussion, I would hope that crats would evaluate consensus partly by looking at the discussion and seeing which arguments were actually resolved in the discussion, and also giving less weight where people reduced the weight of their vote by marking it weak, or "moral support". It is normal in AFD to extend discussion where there has been little input, and RFAs have been usefully extended where new points have arisen late in the day, neither of these arguments applies to Liz's RFA while we have the counter argument of there already having been offsite canvassing. I would agree with extending an RFA where participation was low, but that clearly doesn't apply on the English language wikipedia as our lowest successful !vote this year was 79. ϢereSpielChequers 16:17, 20 August 2015 (UTC)- WSC, he's talking about Cyberpower, not Cyclonebiscuit. Too many C's lately.
- DHeyward, you can't just cherry-pick two examples and say 'well, this is what they look like to me'. Look at all non-unanimous successes and see what you find. So far I see that smaller numbers give you noisier data (!!). Of course, I'm also firmly convinced that closing Cyberpower's RfA as "no consensus" was a crap decision. Opabinia regalis (talk) 16:24, 20 August 2015 (UTC)
- Whoops, my bad, one of my comments struck, the others still apply. As for the crats' decision re Cyberpower678, I'm not sure which way it would have gone if the candidate hadn't withdrawn during the crat chat, but once Cyberpower678 withdrew a no consensus decision seemed right to me. ϢereSpielChequers 18:04, 20 August 2015 (UTC)
- I didn't cherry pick them, rather they were close in time temporally, they both had crat chats and they both had reasonably similar final percentages with opposite outcomes so it begs the question as to why that happened. As for smaller data giving noisier results, I'm not sure how that to handle that. One the one hand, it gave more control of the RfA to smaller group of opposers in Cyberpower's RfA allowing very few !votes to push him into the discretionary zone. One the the other hand, Liz's RfA had a large number of supports which means it had a large number of opposers as well. The overall impact of individual support/oppose !vote diminishes with quantity. It all comes back to the null hypothesis of whether, in the discretionary zone, consensus is assumed or consensus is not assumed. Maybe the null hypothesis is dependent on sample size and trend correlation. But data is the key starting point. --DHeyward (talk) 20:40, 21 August 2015 (UTC)
- Whoops, my bad, one of my comments struck, the others still apply. As for the crats' decision re Cyberpower678, I'm not sure which way it would have gone if the candidate hadn't withdrawn during the crat chat, but once Cyberpower678 withdrew a no consensus decision seemed right to me. ϢereSpielChequers 18:04, 20 August 2015 (UTC)
- The fundamental problem here is that it ignores the qualitative aspect of the process. While I didn't participate in the cyberpower crat chat, the reason why an RFA in crat chat succeeds or fails is based on a subjective consideration, and these are merely but two data points. Remember, if an RFA ends in the discretionary range, a bureaucrat may simply close it according to his or her reading rather than open a crat chat. A crat chat by its nature is a very close case with a very subjective determination. Andrevan@ 18:17, 20 August 2015 (UTC)
- It is subjective but 'crats always couch their decision in "consensus." They also decide "clear consensus" based on hard numbers (i.e. above 80% and below 70% are "clear consensus/no consensus"). It seems odd, then, when they make comments about canvassing, !vote discounting, and then state in a 'crat chat about "consensus" or "no consensus" without even looking at data. there's no evidence in either of the two that I posted regarding canvassing and the front and back end vote patterns look similar. Both were above the 80% discretionary zone shortly before closing (~2 days before for Liz and ~1 day before for CyberPower). It seems odd to me that a decision is purely data driven except when data analysis is needed the most. I chose these because bother were abotu the same in the end yet one succedded and one did not and the questions is "why?" In the discretionary zone, the end number is the least important figure. How 'crats are divining "consensus" in the discretionary zone should have more data before they a) decide to close an RfA and b) decide what consensus is. Even the null hypothesis of consensus/no consensus was either not understood by all 'crats or not resolved prior to discussion. That's a fundamental question that should be agreed to before discussion even starts. Since this is not voting, 'crats shouldn't immediately start looking for hanging chads and other trivialities and, rather, start trying to analyze the will of the community. Some of that requires baseline assumptions and data analyses. Personally, I think the last day vote tails are bad and RfA's shouldn't have hard close dates so the last minute rush doesn't happen. --DHeyward (talk) 00:13, 21 August 2015 (UTC)
- The biggest difference between those two RFAs is that in one case the candidate withdrew during the crat chat and it was closed as no consensus, and the other the candidate did not withdraw and it succeeded. If Liz had withdrawn during the crat chat then obviously her RFA would have been closed as no consensus. If cyberpower had not withdrawn then the crats would have had to make a call one way or the other. ϢereSpielChequers 12:26, 25 August 2015 (UTC)
- That's incorrect. It was closed "No Consensus" instead of "Withdrawn" precisely because the crat chat was "No Consensus". It's easy to follow the crat chat in Cyberpower to see that it was going to fail before he tried to withdraw. He was not allowed to withdraw. They note that in closing. And if that is the only difference between the RfA's, I'd argue that Cyberpower should have received the bit and then give it up if he wished. --DHeyward (talk) 14:48, 25 August 2015 (UTC)
- I didn't say it was the only difference between the RFAs, I merely pointed out that it was the biggest difference. As for the idea that the crats would not allow someone to withdraw, I suggest you read the last few posts of the crat chat As for whether it would otherwise have closed as a success or fail, you may have an opinion, but we don't and can't know. ϢereSpielChequers 17:19, 25 August 2015 (UTC)
- They close withdrawals as "Withdrawn". They close No Consensus as "No Consensus." There is nothing to opine about. It was "No Consensus" and as you have pointed out, the response of the crats was the reason for "No Consensus." --DHeyward (talk) 18:42, 25 August 2015 (UTC)
- Closing as Withdrawn usually means that the candidate withdrew before the 7 days were up. What happened in Cyberpower's case is that he withdrew during the crat chat and the crats closed it as no consensus in immediate response to his withdrawal. That makes it difficult to compare to Liz, and as I explained the biggest difference between those crat chats is that one ended abruptly on the withdrawal of the candidate. ϢereSpielChequers 19:32, 25 August 2015 (UTC)
- The first few crats to comment in the Cyberpower case said they would've promoted without a chat, but it was clearly heading for a 'no consensus' result at the time he withdrew. On the other hand, the initial comments in Liz's case suggested a ' no consensus' outcome and drifted the other way. Both seemed to blow in the winds of the talk-page commentary. While Cyber's withdrawal was a thoughtful decision, IMO the crats set a poor precedent in that one by not publicly discounting opposition of the form "this fellow volunteer isn't prioritizing the things I like, so I oppose in the hopes he'll work on my favorite projects". Opabinia regalis (talk) 20:47, 25 August 2015 (UTC)
- Closing as Withdrawn usually means that the candidate withdrew before the 7 days were up. What happened in Cyberpower's case is that he withdrew during the crat chat and the crats closed it as no consensus in immediate response to his withdrawal. That makes it difficult to compare to Liz, and as I explained the biggest difference between those crat chats is that one ended abruptly on the withdrawal of the candidate. ϢereSpielChequers 19:32, 25 August 2015 (UTC)
- They close withdrawals as "Withdrawn". They close No Consensus as "No Consensus." There is nothing to opine about. It was "No Consensus" and as you have pointed out, the response of the crats was the reason for "No Consensus." --DHeyward (talk) 18:42, 25 August 2015 (UTC)
- I didn't say it was the only difference between the RFAs, I merely pointed out that it was the biggest difference. As for the idea that the crats would not allow someone to withdraw, I suggest you read the last few posts of the crat chat As for whether it would otherwise have closed as a success or fail, you may have an opinion, but we don't and can't know. ϢereSpielChequers 17:19, 25 August 2015 (UTC)
- That's incorrect. It was closed "No Consensus" instead of "Withdrawn" precisely because the crat chat was "No Consensus". It's easy to follow the crat chat in Cyberpower to see that it was going to fail before he tried to withdraw. He was not allowed to withdraw. They note that in closing. And if that is the only difference between the RfA's, I'd argue that Cyberpower should have received the bit and then give it up if he wished. --DHeyward (talk) 14:48, 25 August 2015 (UTC)
- The biggest difference between those two RFAs is that in one case the candidate withdrew during the crat chat and it was closed as no consensus, and the other the candidate did not withdraw and it succeeded. If Liz had withdrawn during the crat chat then obviously her RFA would have been closed as no consensus. If cyberpower had not withdrawn then the crats would have had to make a call one way or the other. ϢereSpielChequers 12:26, 25 August 2015 (UTC)
- It is subjective but 'crats always couch their decision in "consensus." They also decide "clear consensus" based on hard numbers (i.e. above 80% and below 70% are "clear consensus/no consensus"). It seems odd, then, when they make comments about canvassing, !vote discounting, and then state in a 'crat chat about "consensus" or "no consensus" without even looking at data. there's no evidence in either of the two that I posted regarding canvassing and the front and back end vote patterns look similar. Both were above the 80% discretionary zone shortly before closing (~2 days before for Liz and ~1 day before for CyberPower). It seems odd to me that a decision is purely data driven except when data analysis is needed the most. I chose these because bother were abotu the same in the end yet one succedded and one did not and the questions is "why?" In the discretionary zone, the end number is the least important figure. How 'crats are divining "consensus" in the discretionary zone should have more data before they a) decide to close an RfA and b) decide what consensus is. Even the null hypothesis of consensus/no consensus was either not understood by all 'crats or not resolved prior to discussion. That's a fundamental question that should be agreed to before discussion even starts. Since this is not voting, 'crats shouldn't immediately start looking for hanging chads and other trivialities and, rather, start trying to analyze the will of the community. Some of that requires baseline assumptions and data analyses. Personally, I think the last day vote tails are bad and RfA's shouldn't have hard close dates so the last minute rush doesn't happen. --DHeyward (talk) 00:13, 21 August 2015 (UTC)