@22 who did Jimmy piss off in 2007?

@fenwick67 Definitely not the Sinosphere, nor the Arabophone, those grew through that very distinctive U-turn. I was hoping someone could tell me what happened everywhere else.

Maybe the cutest thing is how English, French, and Russian (the European~) lines have the periodic dips at year-end as people celebrate Xmas/New Years by not editing Wikipedia.

Meanwhile, the Japanese, Chinese, Arabic, and Hebrew lines lol dgaf about holidays, they edit Wikipedia without any serious break (at this resolution at least).

Well color me purple, I'd have thought more #Wikipedia editing happens during the weekend—imaging hardworking Wiki editors wrapping up a week of work work, looking forward to the weekend to do some Real Wiki Work—but nop.

Here's the last year+ of human editors (anon+registered) seen on English each day, by day of week.

Mon–Thurs dominate, some slacking on Friday, tons of slacking on Saturday. Sunday is between Saturday and Friday.

O hai Japan.

Japanese editors match my mental model, logging in in bigger droves on the weekend to contribute.

@22 i would like to see the plot with the y-axis running all yhe way to 0 though

@Maltimore Here it is (ignore the drop to 0 at the right, missing data). So ~200 fewer Japanese editors log in during the work-week than the weekends, around 9% (which is similar to the drop between Mon–Thurs and Friday for the English Wiki).

I'll be predicting daily editing activity, so I'm really interested in finding reliable trends to remove from the data before training prediction models, and the day--of-week is a great predictor. I'll post some other cycle analysis in a few hours.

I've spent the last 45 hours, mainly going down blind alleys, but in between that making some really interesting observations (through code and charts) about the periodicities and cycles in #Wikipedia editor activity. Here are the most salient ones (I seem to have put many comments in the accessibility text for each image—will Masodon clip my text to make room?).

I'll add these to the project README this weekend but very briefly:
- daily #Wikipedia editor activity (see past posts in this thread!) for all these seven languages exhibits a long-term positive autocorrelation (its a long-term memory process with a Hurst exponent probably ~1).
- Most languages have a strong weekly and yearly cyclicity.
- However, there are definitely many seasonal cycles (cycles within cycles) that govern autocorrelation.

Predicting these is going to be fun!!

@dulcet Ah, a veteran from those years explained to me that that's when the great war between the inclusionists and the deletionists happened and the latter won—only notable articles allowed, so no more articles for every single character in Pokémon or Romance of the Three Kingdoms. They recommended I read Gwern's report (I love Gwern so I need no excuse!):

https://www.gwern.net/In-Defense-Of-Inclusionism

@icefox a friend who saw it happen told me that was when the war between the inclusionists and the deletionists ended and the latter won: only notable articles allowed, get your filthy articles about Pokémon and such to Wikia:l.

https://www.gwern.net/In-Defense-Of-Inclusionism was recommended to understand what happened. Gwern there talks about how the Wikipedia people saw what happened and sought to staunch the loss but failed. There was a slight uptick starting in mid-2014 that you can just tell in that plot but… 🤕.

@icefox personally I found Gwern’s explanations of stagnant editor and content growth convincing. Sad but convincing. But no doubt there are others with different explanations l.

@22 @icefox tbh deletionism is good

@XavierJulep @icefox trolllol