Best Twitter Filters to avoid following bots

[sexysphere]

Hello,

What could be the best twitter filters to follow real users and avoid following bots?

- Accounts with profile image?
- No url in the profile?
...?

Thank you.

 

[jamie3000]

I coded my bot to skip eggs and run the name against a database of common first names for whatever geo I'm targeting. Works pretty well.

 

[survum]

If you use followliker, check the advanced settings and require a follow count of 200+ and a tweet within the last 30 days. In addition to obviously requiring a profile picture.

 

[Iconoclast]

you can use tweepi

its good

if you have old account then you can easily follow targeted followers by using their settings

 

[punkinhead]

original tweets is big. A lot of bots just retweet if anything. An original tweet within 30 days is pretty good indication.

Banner image is good, but I don't know of bot that has that filter. Most just do profile image. Most bots I see these days have profile pic, tho.

First name thing sounds like a good idea, but obviously will miss a number of real folks. Depends on which side you want to err.

I always try to take a step back, though. People get hung up about this criteria vs that, but at the end of the day, what is it you REALLY want? Other accounts to retweet your stuff to their followings. Whether you show as having a dozen followers or a million, it's the retweets from others that really gets you beyond your immediate sphere of influence. Frankly if an acct does that, I don't care if it's a bot or not... and a real user who has never shared anything from one of my competitors is useless too. Personally, I find more value in those accts that retweet my competition than in any other filter I can think of.

 

[uncce]

I coded my bot to skip eggs and run the name against a database of common first names for whatever geo I'm targeting. Works pretty well.

i had similar idea about checking common names, numbers in usernames, special chars, check for facebook nick etc etc. but the reality is you do not need to do that 'clever' analyzing. the problem is - a LOT of false positives.
what i did at the end is dumped a massive db and looked for patterns in different user stats. and there are clearly visible differences between account types - marketer, bot, hyperactive teenager etc. patterns keep staying constant even when the db is growing. actually the most basic numbers are most important such as account age, location, followers,following and a few more.

 

[jamie3000]

i had similar idea about checking common names, numbers in usernames, special chars, check for facebook nick etc etc. but the reality is you do not need to do that 'clever' analyzing. the problem is - a LOT of false positives.
what i did at the end is dumped a massive db and looked for patterns in different user stats. and there are clearly visible differences between account types - marketer, bot, hyperactive teenager etc. patterns keep staying constant even when the db is growing. actually the most basic numbers are most important such as account age, location, followers,following and a few more.

Is that with the API your pulling those stats?

 

[uncce]

no, API has ridiculous limits. i just dump the source. stuff like tweet count, urls, mentions are just from last 20 tweets. but its good enough. plus i can do multithreading where only limit is my internet connection speed not twitter rules

 

[avalon1981]

This is settings from Twidium Inviter: