Requirements: Not arwiki, huwiki, cswiki (all of which already have this). Not huge wikis (if we use >10K active editors per month, then this excludes enwiki, dewiki, frwiki, jawiki, eswiki, and ruwiki). They need to be big enough to get a medium-to-high use of talk pages. Also, I believe that Peter would prefer the wikis in this test to be all Wikipedias.

Here is a list of 16 Wikipedias that meet all of these requirements:

ptwiki
fawiki
trwiki
idwiki
plwiki
hiwiki
nlwiki
ukwiki
thwiki
azwiki
rowiki
bnwiki
hewiki
kowiki
viwiki
zhwiki

Thanks @Whatamidoing-WMF. This list of 16 wikipedias looks good to me based on the requirements you identified.

I took a look at the number of distinct talk page editors at the proposed wikis and confirmed that they are sufficient to produce a sufficient amount of events as well as provide a diverse look at usage for the analysis. (See stats summary below)

Some stats for 16 proposed wikis (Based on edits logged in March 2023)

• Monthly registered talk page editors: 13,361.
• All wikis have at least 100 registered monthly talk page editors, which is sufficient for per wiki analysis.
• Number of registered monthly talk page Junior Contributors: 2,634.
• Number of registered monthly talk page Junior Contributors per wiki: Ranges from 64 editors on azwiki to 902 editors on ptwiki.
• Appears to be a good representation of people editing from various regions.

Assigning back to you to confirm with the identified Wikipedias.

@MNeisler, I'm assigning this back to you, with a request that you double-check the numbers to see whether this will be enough to produce skin-specific results for Vector 2010 and Vector 2022. At the moment, I believe that plwiki, hiwiki, nlwiki, ukwiki, azwiki, and zhwiki default to Vector 2010, and at the others, it's likely that about 20% of senior contributors have switched back to Vector 2010. If we need to add more to the list, then that's possible.

@Whatamidoing-WMF

I reviewed talk page views on each vector skin based on data logged in desktopwebuiactionstracking to get a sense of the scale of data. Based on those numbers, I think we will have enough data to produce skin-specific results overall for Vector 2010 and Vector 2022.

A couple notes:

• We may have limited data for some of the smaller wikis, especially on the non-default skin. I'd recommend we still include these wikis as they help ensure we have a diverse set of wikis in the test. Similar to what has been done in past analyses, data from these wikis will be included in the overall results but excluded from the per wiki analysis section of the report.
• It looks like all these wikis are sampled at a rate of 20% in DesktopWebUIActions and 100% in EditAttemptStep. We'll look into whether we need to increase the sampling rate of DesktopWebUIActions for these wikis in T330938, which can help provide even more events to review if needed.
• After the test runs for two weeks, I will review the data logged to determine if the test can end or if we need to run longer to collect additional data.
• For the following wikis, the majority of logged talk page sessions in March 2023 were on Vector 2010, which aligns with the default settings you mentioned in T332946#8791169
  • azwiki
  • hiwiki
  • nlwiki
  • plwiki
  • ukwiki
  • zhwiki

Because of the Vector 2022 deployment schedule, I think we should remove plwiki and zhwiki from this list.

If we need more Vector 2010 wikis for the test, we might be able to add dewiki, ruwiki, or itwiki.

@Whatamidoing-WMF I think it would be good to add a couple more Vector 2010 wikis if we can.