User:Rspeer/Editcount inflation

I had a script collect, for all recorded RfAs, the date of the request, the number of edits the candidate was reported to have made at the time, and whether the request was successful. There's some noise in the number of edits, as we only had a standard format for reporting edit counts in 2007 and the summer of 2006; for all other times the script looked for text in the nominator's statement such as "over 2,000 edits" and tried to make sure it wasn't in the context of a particular namespace. (You'll see a couple of low outliers where this didn't quite work.)

The result is most easily seen on a log-scale graph, as edit counts are distributed exponentially, and the edit counts of viable RfA candidates seem to be increasing exponentially as well.



To determine how edit counts are increasing, I fit a best-fit line to the log graph of successful candidacies. The result says that the typical successful RfA candidate's number of edits is multiplied by 1.51 every year, or approximately doubles every 20 months. With the assumption that edit counts are exponentially distributed, the best fit line shows the median of the edit counts of successful candidates. Candidates with less than half the median number of edits are very likely to fail. (I'd like to make this part more formal, by taking the unsuccessful candidacies into account and finding the line where the candidate's chances of success are, say, 25%. But I don't know what kind of statistics to use for this.)

The exponential equation for this line is $$375 e^{.418 t}$$, where t is the number of years since 2000. (Yes, I know there was no Wikipedia in 2000. It's a good reference point to start calculating from.)

According to this model:
 * Current admin candidates are expected to have at least 4000 edits, and the average candidate has 8000.
 * In January 2004 -- just before RfAs started being archived consistently -- candidates were expected to have 1000 edits, and the average candidate had 2000.
 * By January 2008, the average candidate will have 10,000 edits (and candidates with less than 5000 will be shot down).
 * In late 2009, it will be nearly impossible to be promoted with a four-digit edit count, as the median edit count of new admins will have risen to 20,000.
 * Within a few years, it will be difficult to actually evaluate the contributions of most admin candidates, as clicking "last 500" on their contributions page will only show you a couple of days worth of edits.

Editcountitis is increasingly causing a separation between "regular" editors and "admin-track" editors, where "admin-track" editors are the ones who do mindless tasks that produce lots of edits very quickly. Can you help fight editcount inflation by questioning all votes (or !votes, if you must) that are based on edit count?

 r speer  / ɹəəds ɹ  07:16, 9 September 2007 (UTC)