Bitcoin Forum

Updates:

• Followback verification is implemented (https://bitcointalk.org/index.php?topic=4899879.msg44247726#msg44247726)
• Multi-search is implemented (https://bitcointalk.org/index.php?topic=4899879.msg44419532#msg44419532)
• Tweets analysis algorithm is improved (https://bitcointalk.org/index.php?topic=4899879.msg44478595#msg44478595)
• Multi-account detection is implemented (https://bitcointalk.org/index.php?topic=4899879.msg45276328#msg45276328)

Bots. Twitter bots. They are the problem actual for both sides: bounty hunters and managers.
Bounty hunters get paid less and in some cases are not able to enter a campaign (e.g. bountyhive campaigns) due to a large number of participants, most of whom are bots.
Bounty managers also suffer from them as they need to analyze more Twitter accounts and check more reports.


An existing solution (Twitter Audit) is useless in finding bots and analyzes only users’ followers.
Bot accounts have basically follow-for-follow followers who are marked as real by the service, so it does not bear any value.
What’s the solution? TwitterScoring (https://twitterscoring.com). This is a web application that automatically analyzes Twitter profiles and gives them scores depending on their quality.
The score is based on 3 main criteria: personal tweets, posting timings and social engagement.


Personal tweets. It shows how much a specific user posts his own tweets. TwitterScoring analyzes last 100 posts excluding replies. Bot accounts are full of retweets from different projects, so they will not get high scores for this criterion.
Posting timings. It shows how often a specific user posts from his account. Bot accounts participate in many bounty campaigns, so they need to post a lot to meet their rules. Such behavior is noted by TwitterScoring and users with overposting can't get high scores.
Social engagement. It shows how active is the user audience. As already noted, bot accounts have basically follow-for-follow followers who do not really follow their updates, so bot profiles do not have likes and retweets under their tweets and in most cases get 0 for this criterion.

All criteria are calibrated on more than 10 000 real Twitter accounts. It’s not hard to get a score of 50 if you have enough personal tweets and do not post too much. The remaining half is based on social engagement and it’s not an easy task to get a score of 100 in total. TwitterScoring was designed to rank people from nobody to celebrity and that is why social engagement means a lot.


Not only bot accounts can get low scores on TwitterScoring, but also people who prefer quantity over quality as they post too much and often quickly to complete all campaigns at one time. So, how to determine whether a specific user is a bot or not and to not leave behind all bounty hunters?

Analysis of DAOstack bounty participants (report (https://docs.google.com/spreadsheets/d/1Low9rIQMhoHawdvT6jDEPT-O1kXFEbHb6_A10y8aF2Y/edit?usp=sharing)) has shown that users with scores less than 10 can be considered as bots. Such users have 4 personal tweets and a 40-minute difference between posts (36 posts per day) on average.
People with scores more than 10 and less than 20 can be considered as quantitative hunters. They have much more personal tweets (23), but the posting difference is approximately equal to the bots’ value.
Half of the people (566 participants) have a score greater than 27 (update (https://bitcointalk.org/index.php?topic=4899879.msg44478595#msg44478595)). Starting from a score of 30, average personal tweets count increases to 44 and the posting difference goes to 360 minutes (4 posts per day). The score of 30 would be the optimal value for most Twitter campaigns – a significant part of bounty hunters will remain, they will get paid at least twice as much and BMs work will decrease by the same amount.


How to get your personal score in TwitterScoring? (Guides (https://bitcointalk.org/index.php?topic=4899879.msg44575975#msg44575975))
1. Click on the profile icon on the left side of the search bar.
2. Authorize with Twitter.
3. Click on the update button.
4. Wait for your profile to be checked.
5. If everything is OK, you will see “Status: Updated”. The first update is free, after that you will need to pay 0.5$ (in ETH) to update your profile again. Active status remains for 1 month.

If you’d like to check your friends’ or bounty hunters’ profiles, then just type a username in the search bar and submit it. (Multi-search (https://bitcointalk.org/index.php?topic=4899879.msg44419532#msg44419532))

Op asked me to post my personal thoughts on this and I feel glad that someone is coming up with this by providing better alternate of Twitter Audit.

---

This service might work but there are still certain users who uses some follow back tags or a sort for follow for follow exchange. IMO users following in this only follow for the reason that they will gain followers. So basically they are not in for the content they tweet (they have no interest in). So I want to know if you have any sort of thing in mind to counter this?

Also I think rating on the basis of social engagement is pretty good. It would be much more good if you provide another sort of engagement like cryptocurrency engagement of a user profile.

Rest I really find this good. Is there any cost while looking out for multiple bounty participants Twitter profiles rating?

---

Though I don't know about the permission of this app yet. So please check before you authorize (as I only posted above by reading the content in the main thread)

Thanks for you reply.

I've pondered over this and decided to implement checking profiles for following back. Followback accounts will get lower scores and they will be specially marked. This will be released soon.

TwitterScoring is not designed as a crypto related service, the only thing that connects it with crypto is ETH payments. So, the app can be used by usual people not closely bonded with crypto and such criterion like cryptocurrency engagement may not be of interest to them, but something like the most related profile tags may appear, which will show a profile orientation and will not affect the profile score.

No, there is no any cost.
You can check bounty participants' profiles manually using the TwitterScoring search or automatically (soon) through a special page it.

Only read permissions. Nothing suspicious.

How to check multiple users at once?
Just copy your friends' / bounty participants' usernames (from Excel / Google Spreadsheet, for example) and paste them into the TwitterScoring search.
Usernames can be of any format and there is no need to make them of a single format, so your list of usernames may look like this:
As a result you will get a .txt file including the following information:

• Username
• Followback status
  • 0 – Non-Followback
  • 1 – Followback
• Total score
• Profile status
  • 0 – Inactive
  • 1 – Active

To work with the data in a spreadsheet simply copy all the file content into it.

All connected profiles have been automatically updated.
A new analysis of DAOstack bounty participants has shown a downgrade in the total score of many users.
Current average score is 23 (previous - 27).

TwitterScoring is designed as a free service for managers and their only task is to notify bounty hunters about it in rules of Twitter campaigns.
But those managers who do not want to introduce TwitterScoring among their participants can order a report that includes user scores and other related information (you can specify what to add/remove).


Minimum number of Twitter accounts per order is 100. Price per one user is changing from 0.1$ (100 users) to 0.05$ (1000 users). The more you order the less you pay per single account. Starting from 1000 users, the price is fixed at 0.05$ per user. Use the price calculator (https://twitterscoring.com/price_calculator) to estimate the cost of your order. Payments are accepted only in cryptocurrencies.


You can also test the service for free on your Twitter accounts. Send me a message with a list of Twitter usernames/links (up to 3000) to get your report.

What is TwitterScoring?
TwitterScoring (https://twitterscoring.com/) is a web application that analyzes Twitter profiles and gives them scores depending on their quality.

What does TwitterScoring analyze?
TwitterScoring analyzes personal tweets, posting timings (time intervals between user posts) and social engagement (number of favorites/retweets under user posts).

How can I get my personal score? How much does this cost?
Follow the guide (https://bitcointalk.org/index.php?topic=4899879.msg44575975#msg44575975). The first update is free, after that you will need to pay 0.5$ (in ETH) to update your profile again. Active status remains for 1 month.

I am a bounty manager. Why should I use TwitterScoring in my campaigns?
Your campaigns will not have bot and multi-account participants. Other participants will get more tokens for their participation and they will have more trust in you.
You will have fewer bounty participants, but enough to make a project promotion. The fewer participants you have, the less work you do and the fewer complaints you get.
You will get quality Twitter profiles for your campaigns. Twitter promotion will not look like a promotion made by third parties. So, you will get more trust from ICO project managers.

Ok. How can I get the TwitterScoring scores for multiple users?
At first, you need to notify your bounty participants about TwitterScoring. Just add a new rule to your campaign and wait a bit until users update their profiles. Then follow the guide (https://bitcointalk.org/index.php?topic=4899879.msg44575975#msg44575975) to get their scores. This is a free option.

I don't want to introduce TwitterScoring among my bounty participants. Can I get their scores without attracting them?
Yes, you can order a report (https://docs.google.com/spreadsheets/d/1Low9rIQMhoHawdvT6jDEPT-O1kXFEbHb6_A10y8aF2Y/edit?usp=sharing) that includes user scores and other related information. Use the price calculator (https://twitterscoring.com/price_calculator) to estimate the cost of your order.

What is the optimal score for most bounty participants?
The recommended score is 10. Users with scores less than 10 can be considered as bots. Such users have 4 personal tweets and a 40-minute difference between posts (36 posts per day) on average. The score of 50 is the highest possible score for usual people without active audience.

I'd like to test TwitterScoring on my Twitter accounts. Is it possible?
Yes, it's possible. Send me a message (https://t.me/quesz) with a list of Twitter usernames/links to get a free report.