User talk:Rspeer/Editcount inflation

Hey rspeer. Looking back at the comments I made on WT:RFA, I hope I wasn't too harsh by using the term "bad science". It is very interesting data but I think there are a couple of problems with your analysis. All in all, I'm not convinced by the doomsday predictions. That's not to say that editcountitis shouldn't be fought against. It should as should all ludicrous opposition on RfA, the "no need for tools", the "thinks that RfA is a vote", the "no experience with images", the "no vandal fighting", the "too much vandal fighting". Pascal.Tesson 23:35, 10 September 2007 (UTC)
 * By fitting a line to the log graph, you are in effect assuming that the threshold is increasing exponentially. I don't think it makes sense to project that growth on a two year horizon.
 * The analysis of the data assumes that editcount is the significant factor. From our own experiences, we know that it is significant, both positively and negatively but to infer from the graph itself that admin candidates are expected to have 4K edits is a stretch.
 * It would be interesting to compute the percentage of RfAs that succeed only the second time around but I would think it is significant enough to skew the data.
 * It's clear that you want your data to say that editcount has run amok but in the process you're throwing away any possibility for alternate explanations of the phenomenon. For instance, regardless of editcount, there has been a growing trend of asking candidates to have experience with project space, images, vandal fighting, article writing and the deletion process. Failing on any of these counts will kill your RfA. This pushes up the average editcount but this has little to do with editcountitis. Yes, there are increasing demands for admins and that pushes the editcount up. But that does not make editcountitis the explanation of that increase. Also, the minimum time spent on the project is also going up and all the user scripts have unquestionably increased the rate at which the typical editor is amassing edits. This is even more true for Wikipedia regulars who are the ones most likely to be using scripts.
 * Saying that RfA is getting skewed towards what you call admin-track users ignores the reality of what's actually happening on RfA. AWB edits are being routinely discarded as junk.
 * In the real world, filled with negligent human beings, mathematical modelling does not always work indefinitely. For example, Mark Twain famously predicted the Mississippi River would shrink to a mile in length.  I've actually been doing OR on this for years, but of course have gotten none of it published.  From anecdotal evidence, I think the magic number for edit counts has settled at about 3,000.  More edits than that, the !vote has hovered between 75 and 100%.  Users with fewer than that can convince others that they have made significant contributions - say creating images or especially tricky articles, but often still reach a majority of 50 %.  That being said, I safely waited until I hit the 6,000 mark. Bearian 16:38, 21 September 2007 (UTC)
 * You've been doing research on edit counts, Mark Twain, or the shrinking Mississippi? :) I suppose you mean edit counts. I have seen a bit of a turnaround in opinion about edit counts recently, and there was the promising case of Earle Martin (though I think the unique tone of his nomination got everyone who was frustrated with RfA to stop lurking and support him). I can hope that a qualified person with under 3,000 edits can still be promoted next year, but from a pessimistic and statistical point of view I doubt it will happen.  r speer  / ɹəəds ɹ  17:45, 21 September 2007 (UTC)
 * Note that my own optimism is being shattered by some of the comments at Requests for adminship/Ronnotel. Perhaps you are closer to the truth than I want to admit! It now seems that 2600 edits over two years is being "a bit wet behind the ears" or, worse yet, the sign that a user has "barely contributed to the project". The madness of it all... Pascal.Tesson 18:51, 21 September 2007 (UTC)

This is interesting, but with any correlation it is possible that a third unincluded co-variable is interfering with your analysis. I recommend repeating this analysis with "time since account was registered" vs success I predict this will be a flat line. Then take a random sample of wikipedians at different years and compare the median numbers of edits normalised to time - edits per week for example. I predict this will be an exponential curve.

My hypothesis is that we are looking at a consequence of general edit inflation, probably the result of automated tools such as twinkle and AWB, with the time that candidates have been around remaining constant, but the productivity of all editors increasing. Tim Vickers 15:26, 22 September 2007 (UTC)