I think the best way to track would be to select a random group of 1000 active editor vets each year (who have been editing for 2+ years and are thus probably going to be around for at least a year more) and track their edits during the course of a year, and then based on the same criteria, the next year make a new random selection, etc.
I would assume that over time, the yearly activities of such a random group will show some interesting seasonal fluctuations based on the language (and thus the holidays in the country of origin) of such editors.
Basing your data on actual usernames that have been somehow vetted to be active editors and not temporary bot- or project-related would be easier than just eliminating all bots.
I believe a similar study was done for new contributors, but I don't know how the selection was done.
2013/7/23, Ryan Kaldari rkaldari@wikimedia.org:
I was poking around on stats.wikimedia.org and reportcard.wmflabs.org to see if I could find out how overall editing levels had changed (if any) over the past year. Unfortunately, it seems that all of our "edits per month" graphs show all edits, including bot edits. Since changes in bot editing levels are often dramatic from month to month, this noise effectively cancels out the usefulness of the graphs. For example, you can see a huge spike in March when I presume the Wikidata bots were running at full force: http://reportcard.wmflabs.org/#secondary-graphs-tab http://stats.wikimedia.org/EN/ChartsWikipediaEN.htm#3
My question is: Would it be possible to replace or augment these graphs with graphs that exclude bot edits? I know that bot status is not stored in the revision table, so this would be quite expensive to tally. Would it be prohibitively expensive? Sorry if this is a dumb question.
Ryan Kaldari
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics