Wikipedia:RFA study

This page aims to list not just every RFA that happened this year so far or the four years before, but also lots of characteristics about them and the candidates therein, to try to see if there are patterns that predict whether the RFA will be successful or not.

General trends
These trends are based on data from 2004 to 2018*, taken from WP:SRFA and WP:RFAU. In total, there have been 4,418 RFAs during this time period.

The annual number of total RFAs here on en.wiki peaked in 2007 at 920; the lowest number of RFAs in a given year thus far was 36 in 2016. Currently there have only been 13 RFAs this year, suggesting that this year will probably have the fewest RFAs of any year in en.wiki's history (barring a remarkably rapid blitz of RFAs in the last four months of the year).

The highest success rate for RFAs was in 2004, the first year for which data exists; that year, 79% of RFAs were successful. By contrast, the lowest success rate was in 2012, when only 29% of RFAs were successful. The most recent success rate, for 2018 so far, is a somewhat-less-discouraging 46%. Interestingly, the average success rate for RFAs over all years is 45%.

From these statistics two things are obvious, as many others have already pointed out:
 * 1) There are now way fewer RFAs than there were in the (relatively) distant past, and
 * 2) RFAs are now much less likely to succeed than they were in the past.

* This is based on all RFAs that have closed so far this year.

Accounting for deleted RFAs
One significant limitation to accurately estimating the number of RFAs in a given year is that for the past few years or so, admins have frequently deleted RFAs that were created by very-new users in clear violation of WP:NOTNOW. As noted on the page WP:RFA by month, "By 2015 admins had started deleting "NotNow" RFAs which artificially reduces the unsuccessful figure." So this means that for at least the past 3 or 4 years, the success rate of RFAs based only on the data below will be higher than the real success rate, since this page currently excludes RFAs that were unsuccessful if they were deleted, as such RFAs aren't preserved at WP:RFAU.

The most logical way to try to solve this problem (at least for non-admins) seems to be to look at the history of the page WP:RFA to see when someone edits it to add an RFA that has since been deleted. Then you can count up all the instances when new users do this, only to be reverted quickly by more experienced users. Doing so reveals, first, that new users adding (or at least trying to add) their newly created RFAs to WP:RFA became much less common right after indefinitely extended-confirmed protected the page in November 2017, ending a more than 5-year period of continuous semi-protection. By looking at the WP:RFA page history in this way, I have estimated that there were 4 failed deleted RFAs (FDRFAs) this year so far, and that there were 19 FDRFAs in 2017. Adding the 19 FDRFAs to the total number of RFAs for 2017 lowers the success rate from 21/40 = 53% to 21/59 = 36%. I have also identified 13 FDRFAs in 2016; adding them to the total RFA count for 2016 lowers the success rate for that year from 44% to 33%.

However, this approach involves yet another questionable assumption: that all of the users who tried to add an RFA for themselves genuinely wanted to become an admin, and made their edits to the page in good faith. But this assumption seems to be false pretty frequently: of the 36 users who attempted to add their now-deleted RFA to the main page in 2016, 2017, or (so far) 2018, 22 of them (61%) have since been blocked indefinitely. (Note: this is only based on the ones I included in the above counts; it doesn't include obvious trolling or vandalism, just edits that were clearly attempts to create RFAs.)

2017
*** This is a bit too high as it includes all the edits made in the month in which they were nominated.
 * †This is a bit too low as it doesn't include any of the edits in the month in which they were nominated.

Methods
I determined the amount of time a user had been on Wikipedia at the time of their RFA based on the date of their first edit (as determined from this tool) and the date in the signature in the nominator statement. In the case of Jakec, he was active under the account King jakob c before switching to his current account. Similarly, Liz used the account Nwjerseyliz before switching to her current account. If the nomination was accepted (which defines the start of an RFA) at least a few days closer to the end of a month than to the beginning I included that month's edit count, but if the acceptance occurred at least a few days closer to the beginning, I did not include that month's edits. Edit counts, like the date of first edits, were determined using the Edit Counter tool on XTools.

Time editing was calculated using this tool and the dates described above. The number of days was then divided by 365.25 to get the number of years.

*Time is listed in years, months, unless less than 1 year, in which case, months. Also, if # of months in addition to years is <1, only years will be included.

** These edit counts were estimated by averaging out the difference between edit counts not counting the month of the RFA and counting the month of the RFA, because the RFA did not occur very close to the end or beginning of a month and, in some cases, there was no talk page edit stats summary available.

† This user had admitted to engaging in sockpuppetry, which was why he resigned his position as an admin.

Results from 2015

 * 59 total RFAs in 2015.
 * Overall, 21 (36%) were successful and 38 (64%) were not.
 * 22 self-nominations, 20 nominated by other users.
 * Of the self-nominations, only 5 (23%) were successful. In contrast, of the 20 RFAs in which the candidate was nominated by another user, 10 (50%) were successful.
 * 30 RFAs were first-time nominations, 12 were not.
 * Of the 12 non-first-time nominations, four (SarekOfVulcan, Abecedare, Rich Farmbrough, and Opabinia regalis) had passed RFA before. However, Rich Farmbrough (unlike the other three) was desysopped by ArbCom in 2012. Of the other three, two (Abecedare and Opabinia regalis) lost their adminship status due to inactivity, while the other (SarekOfVulcan) resigned as an admin in 2013. The remaining eight had never been administrators before.
 * Also, of these 12, half were successful and half were not. The proportion of the remaining 28 that were successful was very similar (15/28 = about 54%).
 * In 2015, so far, there have been 5 users who have run for adminship who had previously been blocked (not counting self-requested or accidental blocks). Of these only 1 (SarekofVulcan) was successful.

Edit counts

 * The average edit count of the 15 successful RFAs this year so far is just above 110,000. However, this is skewed considerably by Ser Amantio di Nicolao's edit count of over 1 million edits. Excluding his edit count, the average drops to about 36,000 edits.
 * The standard deviation of edit count among successful candidacies in 2015 is over 285,000 when including Ser Amantio di Nicolao and about 21,000 when excluding him.
 * Overall, including both 2014 and 2015 data, the mean edit count of successful RFAs is about 78,000, and the standard deviation is about 209,000.
 * The mean edit count of unsuccessful RFAs in these two years was higher than that for successful RFAs: about 96,000. The standard deviation was about 312,000.
 * Excluding Ser Amantio di Nicolao, the average edit count of successful RFAs is much lower: about 40,000.
 * With respect to the unsuccessful adminship candidacies, there were three outliers: both of Koavf's RFAs and that of Rich Farmbrough. At the time of his RFA, Rich Farmbrough had nearly a million edits, and Koavf had over a million at the time of both of his. Excluding them, the average for unsuccessful RFAs drops even more, to about 16,000.
 * With respect to median edit counts, the median edit count for successful RFAs in 2014-2015 was about 31,000, and that for unsuccessful ones during the same time period was about 6,000.

Results from 2016

 * Avg. EC for successful RFAs: 36400
 * For unsuccessful ones that number is 18224
 * 9/19 of unsuccessful RFAs were self-noms as compared with none of the successful ones.
 * All 9 self-noms were unsuccessful compared with 16/26 of non-self-noms.
 * Three RFA candidates had been previously blocked; two of their RFAs were successful.
 * 16/19 (84%) of unsuccessful RFAs were the candidate's first, as were 14/16 (88%) of successful ones.
 * Of the 30 first-time RFAs, 14 (47%) were successful; whereas of the 5 non-first-time RFAs, 2 (40%) were successful.
 * Avg. # of years editing for successful RFAs: 7.35
 * For unsuccessful RFAs this figure was 4.38

Results from 2017

 * Average EC for successful RFAs: 50365
 * For unsuccessful RFAs that number is 42258
 * Average # years editing for successful RFAs: 7.8
 * For unsuccessful RFAs that number is 6.9
 * Of the 21 successful RFAs in 2017, we again see a clear pattern: only 2 of them (10%) were self-noms. In contrast, 8 of the 19 unsuccessful RFAs in 2017 (42%) were self-noms.
 * Looking at this issue the other way, of the 10 self-noms, only 2 (20%; and yes, these are the same 2 as above) were successful. In contrast, of the 30 non-self-nom RFAs, 17 (57%) were successful.
 * 14 out of the 33 (42%) first-time RFAs were successful, compared with 5/7 (71%) of the RFAs by candidates who had done one before.