…you may want to have a look at http://twampoline.com first. I’ll wait…

A while back I started wondering about my twitter friends. Specifically how my friends talked to each other and whether there we cliques of friends.

I wasn’t interested in friend/following relationships as that seemed a fairly weak kind of twitter relationship. I was more interested in the “@” stuff that happens in twitter.

I’m a software developer. On Tuesday night I was bored, and it seemed a fairly easy question to answer using technology so I knocked up a quick web app to explore it.

Initially I looked at all mentions and RTs. I started to post some screenshots when the mathematician @PeterRowlett replied to me expressing interest in the graphs.  He hinted that he thought RT/mention relationships was too weak.  Of course, he was right, and now twampoline only looks at direct message relationships.

This is my graph from today.  It’s interesting to see how the graphs change even over a few days. Friends are added, removed, conversations arise (and some drop out of the results) and these all change the graph.  So if you twampoline AdamJTP today you may see a different result.

You can try it yourself by going to http://twampoline.com and entering a twitter username.

…anyway… Here’s the same graph with some explanation.

It’s interesting to see the separation between the islands of friends. 

I’m surprised to see my “maths” friends isolated from my work colleagues and social friends (despite repeated #FollowFridays!).

My “chatting” friends are connected by the social hub @snapplynn without her those groups break apart.

The “local” group seem to be well connected.  This may be because we’re a rural area and there are so few of us on twitter and we tend to huddle together? (the Shropshire twibe has very few members (107 today) considering it represents a whole county).

Now compare @deliciouslunch, who (at time of writing) had a similar number of friends to me. 

Delicious lunch deliver sandwiches locally. It appears that their local network means that there’s a lot of messaging going on with @ShropStar (local paper), @JimInTheMorning (local radio presenter) and @HoptonHouseBnB (Salopian guest house) being quite chatty.  Even without these bigger local stars there’s lots of messaging going on connecting everybody.

How is the map calculated?

From a username it retrieves all their friends.

For each friend it gets the most recent messages to that person (I leave the number of msgs up to the twitter api).

If the message is from another friend then a connection is added (adding the members if they’re not already on there).

Notes:

A message in either direction counts as a link.

I don’t show friends that aren’t in the messaging network.

I don’t add the original user (because I’m looking for islands).

My BIG grand unified theory of twitter messaging networks

I haven’t got one.  There’s no serious point to this.  This is just a toy to satisfy my idle curiosity.

Give it a go yourself on http://twampoline.com and let me know what you think by sending a twitter message to @AdamJTP

Credit where it’s due

Orbifold wrote the clever Silverlight control that moves things about.  I just hooked it into twitter.

@PeterRowlett early tweets with me stopped it being a mess of links (by suggesting removing RTs and mentions)

@snapplynn and @marnweeks for putting up with my repeated following/unfollowing to see how my twitter timeline changed when I was originally thinking about the problem.