Threshhold analysis from: http://localhost:8888/notebooks/determine%20bot%20thresholds.ipynb

# to get 150 per day/ 165 to have some buffer and for control group
dfss.groupby('user_registration_day').agg(lambda df: threshold_to_get_top_n(df,160,'pred_min'))['user_id']

user_registration_day
2018-11-30 Fri   0.884070
2018-12-01 Sat   0.842765
2018-12-02 Sun   0.865676
2018-12-03 Mon   0.906331
2018-12-04 Tue   0.907539
2018-12-05 Wed   0.886038

How did I miss Thursday? OK...Tired

AT worst cut everything at 0.84 Otherwise these 'pred_mins'

Using min instead of mean, is better because it's A) stricter and only 30% of users end up going multi-session

updates:

• using a local mysql database
• using a (dev with ssh tunnel)

still todo:

• exponential backoff for each stage in hostbot_ai.py - https://github.com/litl/backoff
• wednesday - write send_invite
• how does host-bot actually post to wiki, can it be done with mwapi?
• thusday run with RQ
• friday - set up on wikidump parse
• friday - optimize new user query to utilize user-index
• setup mysql with files in actual disk, not nfs
• saturday - launch 2-week experiment to run but printing to test
• setup airbrake?

@awight: What's the status of this work?

I've only had very general conversations about this project so far, and didn't realize there was working code :-)

@notconfusing Would you mind dropping a link to the bot repo, and letting us know whether it's something you might pursue, or if we should pick up the thread?

@notconfusing, do you have a repo for this that you could link us to? I think we could take it from there if you are busy with other stuff.