In order to integrate Revert Risk on the Recent Changes feed, we need to define the score thresholds for labeling revisions as "Likely to be reverted". To set this thresholds we need:

• Check the Precision-Curve recall
• Get the amount of revisions (and confusion matrix) marked as "likely to be reverted" using different thresholds, and compare with the current ORES-based labels.

Ideally, we would like to have one unique threshold for all Wikis.