Wikipedia:WikiProject User warnings/Testing/SDPatrolBot

Kingpin13's SDPatrolBot is potentially important as a tool for reaching new registered editors acting in good faith because its main purpose is interacting with editors of articles tagged for speedy deletion, and it provides warnings to new editors who don't yet understand that removing a CSD template is not the right way to object to speedy deletion.

Warnings
SDPatrolBot delivers several warnings, depending on the experience level of the editor who removes the CSD tag. The ones being test are:


 * 1) User:SDPatrolBot/warning, which is the default
 * 2) User:SDPatrolBot/warning/experienced, which is for non-newbies
 * 3) User:SDPatrolBot/warning/alt, which is the new version for newbies written by Kingpin13, Maryana, and Steven Walling

Testing and analysis
Rather than using our more complex tracking system for Huggle and Twinkle, this instance is much simpler since all contribs are made under a single username (the bot's). We're tracking which templates are delivered to whom through the bot's source, rather than hidden Z templates.

The data from the test so far can be found at:


 * User:SDPatrolBot/AltWarningResults/Raw, for the raw lists of who received the default and alt templates
 * User:SDPatrolBot/AltWarningResults, which includes timestamps and whether the user was reported at AIV

We'd like to measure...


 * 1) How many editors were warned more than once from each group.
 * 2) The edit counts of all users before and after being warned, measured in increments such as: any time, 30 days later, 60 days later.
 * 3) Whether the articles created by the editors were deleted or not.
 * 4) How many editors were blocked from each group.
 * 5) How many editors left a comment on the article talk page.

To Do
Things to do after the preliminary results below.


 * Look at edits and edit summaries by those who were warned again (meaning they removed the speedy tag again).
 * Check point one above to make sure it's not due to the configuration of the bot or something else controllable.
 * Keep running the test to see how the sample changes.
 * Measure warnings other than by SDPatrolBot.
 * Look more closely at the edits by those who were warned a second time and get some qualitative assessment.

Results
Results for all tests are on Meta.