About
trst.me measures user reputation in a way far more robust than counting the number of followers. The basic idea is to look at how many people interact with you and give you their attention, weighted by how many people interact and pay attention to them. The trst.me score is currently based on a scale of 0-10, where 10 is the highest reputation possible.
How ranking works
Imagine if we gave everyone on twitter a bag of jelly beans. Every ten minutes, users hand out all their jelly beans to the people they follow, distributed equally. Meanwhile, their bag refills with jelly beans from all their followers. (It's OK if you eat one or two jelly beans, just not so many it screws up the counting). You'll end up with lots of jelly beans if you have many followers: lots of people hand you jelly beans; if your followers also have high trstrank: they'll have a lot of jelly beans to hand off; and if your followers use discretion about whom they follow: you'll get a good portion of the jelly beans they pass out.
Want more math? We use a 30-machine cluster to produce an iterative eigenvalue solution of the sparse matrix modeling steady-state network flow on the interest graph: 50M+ users, 1.5B+ connections. Want a tad less math? It's similar to the algorithm Google uses to rank websites. From the Wikipedia article on PageRank:
Mathematical PageRanks (out of 100) for a simple network (PageRanks reported by Google are rescaled logarithmically). Page C has a higher PageRank than Page E, even though it has fewer links to it; the link it has is of a much higher value. A web surfer who chooses a random link on every page (but with 15% likelihood jumps to a random page on the whole web) is going to be on Page E for 8.1% of the time. (The 15% likelihood of jumping to an arbitrary page corresponds to a damping factor of 85%.) Without damping, all web surfers would eventually end up on Pages A, B, or C, and all other pages would have PageRank zero. Page A is assumed to link to all pages in the web, because it has no outgoing links.
Why is this different than the number of followers a user has?
Some users are more valuable followers than others. For example:- 20 followers to your account who have 0 followers each are much less valuable than just 1 follower who has 5 followers who each have an average amount of followers themselves.
- If @conanobrien randomly decides one day to follow you, your trstrank will increase much more than if your twitterverse-unknown aunt Marge followed you.
Compare
Our best example of how our results differ is the top users section. Compare our top 10 users with the top 10 users on other Twitter ranking websites:
Full Size
Why is this important?
The importance and trustworthiness of a Twitter user cannot accurately be measured by the number of followers alone. There are many spam accounts on Twitter that create networks of their own users to make it appear as though they are legitimate (ex: user A follows user B and user C. User B follows A and C. User C follows A and B). Ranking users using this algorithm can solve this and many other importance/trust problems.
What are your future plans?
We plan to expand the measure of rank to include other metrics such as the number of @ mentions, retweets, replies and other factors that are relevant to a user's importance and trustworthiness.
FAQ
I entered my username but didn't get a result.
Our data set is fairly comprehensive, but it does not include every single user.
I am a researcher/academic/marketer, can I access the full data set?
Check out what's here: Infochimps' Twitter Datasets.
Who made this?
Infochimps, a site to find data, stats and databases on any subject, made this application by using data from the Twitter trstrank data set.
Where did this data come from?
We've been collecting data from the Twitter API since 2008, and have tweets going back as far as March 2006. Our entire data set contains nearly 50 million users, 7 billion tweets, and over 1 billion connections between users.
Is there an API?
Yes! Learn more here
