While we wait for Jason to iron out the bugs in his TunkRank implementation, I’ve been thinking about the relationship between TunkRank and retweet rank as influence measures.
Here’s my thought: TunkRank assumes in its model that, if X reads a tweet from Y, then there’s a constant probability p that X will retweet it. If this assumption holds true, then the TunkRank of X should be roughly proportional to the retweet rank.
Of course, one of the reasons this assumption might fail is that X is using bots (or bot-like people) to game his or her retweet rank. It’s also possible that the TunkRank assumption about a constant probability of retweeting is too simplistic.
But I’m intrigued at the idea that, subject to the assumptions of its model, TunkRank acts as a sort of ungameable retweet rank.
While we wait for Jason to iron out the bugs in his TunkRank implementation, I’ve been thinking about the relationship between TunkRank and retweet rank as influence measures.
Here’s my thought: TunkRank assumes in its model that, if X reads a tweet from Y, then there’s a constant probability p that X will retweet it. If this assumption holds true, then the TunkRank of X should be roughly proportional to the retweet rank.
Of course, one of the reasons this assumption might fail is that X is using bots (or bot-like people) to game his or her retweet rank. It’s also possible that the TunkRank assumption about a constant probability of retweeting is too simplistic.
But I’m intrigued at the idea that, subject to the assumptions of its model, TunkRank acts as a sort of ungameable retweet rank.