Hi,
I'm getting my Twitter accounts locked every 2-3 days now. Perhaps hard to avoid locks at all, but captchas are getting too hard, need to try and minimize it. What anti-detection patterns pay off best? Would love to hear what others think.
My situation: 20 one year old manually created Twitter accounts. Run on same machine, on own Chrome instances with different proxy locations. Using Windscribe browser extension (all privacy options enabled) + Canvas Fingerprint Defender browser extension. Botting with own developed browser extension.
Own bot characteristics: Twitter actions (follow + like + retweet + comment + post) done with simulated mouse clicks and text entry with chrome.debugger to make mouse/input events isTrusted property be true. Mouse clicks have real-ish random delay between MouseDown and MouseUp events. Click coordinates are random within clickable element. Random delays between actions.
Botting scenarios: I bot Alphabot.app and other raffles where each raffle require 1-3 follows, 1-3 like + retweets. Also Twitter raids where one tweet is liked + retweeted + commented. Typically do 4 bot accounts at once, wait 2-10 minutes, do next 4, etc.
This have worked well for 6 months. Last couple months I have gotten more frequent, and more hard to solve captchas.
I am well aware that my bots behaviour is easy to detect as bots. But not sure where to best put time into making them appear less botish. If possible at all? Could spend 40 hours on one thing and it would make little effect.
Anyone have own experience on how to combat this best?
Here an incomplete list I can think of myself:

I'm getting my Twitter accounts locked every 2-3 days now. Perhaps hard to avoid locks at all, but captchas are getting too hard, need to try and minimize it. What anti-detection patterns pay off best? Would love to hear what others think.
My situation: 20 one year old manually created Twitter accounts. Run on same machine, on own Chrome instances with different proxy locations. Using Windscribe browser extension (all privacy options enabled) + Canvas Fingerprint Defender browser extension. Botting with own developed browser extension.
Own bot characteristics: Twitter actions (follow + like + retweet + comment + post) done with simulated mouse clicks and text entry with chrome.debugger to make mouse/input events isTrusted property be true. Mouse clicks have real-ish random delay between MouseDown and MouseUp events. Click coordinates are random within clickable element. Random delays between actions.
Botting scenarios: I bot Alphabot.app and other raffles where each raffle require 1-3 follows, 1-3 like + retweets. Also Twitter raids where one tweet is liked + retweeted + commented. Typically do 4 bot accounts at once, wait 2-10 minutes, do next 4, etc.
This have worked well for 6 months. Last couple months I have gotten more frequent, and more hard to solve captchas.
I am well aware that my bots behaviour is easy to detect as bots. But not sure where to best put time into making them appear less botish. If possible at all? Could spend 40 hours on one thing and it would make little effect.
Anyone have own experience on how to combat this best?
Here an incomplete list I can think of myself:
- Randomize order of Twitter actions (easy, will do that right away).
- Put in random actions so all bot accounts do not have 100% same actions done.
- Put in random tweets (no own tweets done previusly).
- Manually unfollow random accounts on each bot account?
- Try to get (more) followers on each account?
- Better proxies? Not sure how to do that with Chrome instances.