Summary
jawiki and zhwiki both have FancyCaptcha enabled on edit and create actions for users without skipcaptcha right. This makes these wikis better suited for an A/B test with hCaptcha's 99.9% passive mode. In this task, we'll set up an A/B test (similar to T405239) where 90% of edge unique traffic gets hCaptcha 99.9% passive mode, and 10% of traffic gets FancyCaptcha.
Technical notes
We will need to create an event in the visualEditorFeatureUse event logging stream to indicate whether a given editing session ID is in the experiment or control group. For example:
const experiment = mw.xLab.getExperiment( 'fy25-26-we-4-2-hcaptcha-editing' ); mw.track( 'visualEditorFeatureUse', { feature: 'hcaptcha_experiment', action: experiment.isAssignedGroup( 'control', 'control-2') ? 'FancyCaptcha' : 'hCaptcha' } );
and that could be triggered via mw.trackSubscribe('editAttemptStep'), firing on the loaded event (h/t @DLynch)
Experiment in MPIC: https://mpic.wikimedia.org/experiment/fy25-26-we-4-2-hcaptcha-editing
Acceptance criteria
- 90% of edge unique traffic gets hCaptcha in 99.9% passive mode on jawiki/zhwiki, 10% gets FancyCaptcha on edit/create actions
- VEFU is updated to have an event to track which experiment group a user is in