Essentially this is a bot that tries to detect streams with fake viewers via the Twitch API and the power of statistics.
File layout (primarily):
bdb.py -> twitch_chatters -> handle_twitter
|
v
twitch_viewers
Primarily, it looks at certain metrics such as the ratio between chatters to reported viewers. More bot detection methods are listed below or found in the source code (or the dev branch (which may or may not be 100% up to date)). Let me know if you see any other patterns in streams with fake viewers that i'm not taking into account by emailing me, or feel free to contact me with questions/anything at that email address.
I'm not necessarily accusing the streamers of botting their own channels. I'm not affiliated with Twitch.
"python bdb.py"
Note: To run it locally (not send out tweets), set "tweetmode" to False in
global_consts.py.
Modules needed:
If tweetmode,
Some other detection methods:
New heuristics for detecting chatbots: very low average follower count in
chat (regarding how many people each user is following) is indicative of
chatbots. Also if a lot of the bots have the exact same number (say, under 5).
Or, if most of the followers are following the exact same streams.
The main problems with these is that I need to individually access the
Twitch servers for each chatter; this could take hours for certain streams
with large viewer counts.
Go to http://www.twitter.com/BotDetectorBot!
Program usage: "python bdb.py"