The first thing you have to know is that I am somewhat addicted to statistics. Especially if I am involved at all. So of course statistics for things like my blogs, Facebook, online forums and Twitter take up too much of my personal time. Lately I’ve been looking a lot at Twitter statistics. basically I’m finding them not all that useful. The two big reasons for this is that most of them are opt-in and that the Twitter API limits how much information one can get at a time.

By opt-in I mean that people have to tell the grading services that there is an account that they should  be looking at. TwitterCounter has words on its page that suggest it goes looking for people. Could be but I’m sure they are not finding everyone. So if no one involved with a twitter account tells the service about it then it doesn’t get counted. I suspect that really serious twitter users know about these services and check in on them so it may be that under reporting leans to people who don’t use Twitter that much. Also since one can’t get as much information as one might like on a timely basis there is a long lag time and ratings can easily get out of sync with reality.

But Let’s look at a couple of rating sites.

Twitterholic uses raw follower count to rate Twitter users. The more followers the higher your rating (the lower the rank number). Twitterholic also rates users for locations. This is hard because people are not entering their locations in a consistent format. But they try.

TwitterCounter also just uses follower count for there ranking. If you look at the chart below you will find that my account has a much lower ranking at Twitter Counter than it does at Twitterholic. Clearly Twitter Counter tracks more people. Twitter Counter promises Top 100 lists by location and they seem to usually work for the top 100. Beyond that not so much. There don’t break things down by states but by time zone. A couple of US states are their own time zone though which makes that interesting in a way. On the other hand people are entering time zones in their profiles that have little or no relationship with their actual location. We begin to really understand the problems with location data in Twitter accounts now.

Twinfluence and Twitter.Grader both take the statistics a step or two further. Twinfluence takes into account the number of followers ones followers have. This results in a reach number – the theoretical number of people (ignoring duplicate followers) that a twitter message could reach if every follower sent that message on to their followers. So people whose followers have a lot of followers score higher than those who have a lot of followers who don’t themselves have many followers. Did I lose you yet?

Twitter.Grader uses a similar scheme where the value of your followers is a big part of your own score. This just begs for a recursive mess but it seems to work. Sort of. The thing that stands out here is that my position in the list (raw score) is better in Twinfluence but the percentile I fit into is higher in Twitter.Grader. I’m not sure how to parse that except that Twinfluence probably knows a lot fewer Twitter accounts and the way they rate people is different from Twitter.Grader.

Speaking of other effects of different users being followed. You will notice that Twitterholic rates me as that 13th most followed twitter account but Twitter.Grader ranks me at 24th. This is not because of the way Twitter.Grader ranks people so much as that they know a lot more Twitter accounts than Twitterholic.  All of the people Twitter.Grader shows as higher ranking in the state than me have higher follower counts. Several NH Twitter users who have lower Twitter.Grader scores also have higher follower counts. So the value of the Twitterholic count is marginal in my opinion.

I think these rankings may be of some use for comparison within the tools but not across the tools. Also I think as absolute values they are limited for anything beyond amusement purposes. I’m always looking for other Twitter rating/ranking services BTW so if you know of any drop a comment or Twitter me @alfredtwo.

Now twitter could create good rankings if they wanted to.. They have much easier, faster and more complete access to the data. It could be as simple as rating by followers of course or as complicated as they want. It’s probably an open question as it if there is a good reason for them to do it though. There is plenty of gaming of the system as it is and better ratings might make that worse.