Over the weekend, everyone who is an avid follower of Twitter Grader, noticed a significant change in the algorithm. In my original post I talked about what I noticed to influence the algorithm for twitter grader most heavily. I was quite surprised to see a comment from Dharmesh Shah, the developer for twitter.grader.com, who apparently read my post and said “it’ll be interesting to reconduct the analysis once we go into beta”.
Well, I don’t know if they’re quite in beta yet, but I know that my grade was at a 90 over the weekend, and now its at an 84! So there’s got to be a reason! Here’s what I think…
You will all remember my friend grouping from the last post? Here they are with updated numbers:
These numbers are all as of this morning 11–3, between 11:15 & 11:30 PST.
And just for reference, here is the original list from the last post:
So to start, we can see that there has been quite a bit of variation on the user accounts and on the scoring for almost all users. At the top end, we now have multiple users in the 99th percentile, with several other close to follow. And at the bottom end, scores went from being in the high 50’s — low 60’s to now being high seventies to low to mid 80’s. This is obviously just a result of the user group being an extremely active sampling from twitter. Scores are bound to go up.
But, there are a few things to note that I think are interesting. For example, take a look at davesnyder, his following count went down, follower count went up, and he saw a significant grade change after the algorithm update. -> Clue #1
Clue #2: AussieWebmaster, interestingly enough, saw NO ACTIVITY on his account except for an increase in followers (can you tell me how you did that?), and sees a grade change of almost a full 2 points!
What does this sound like? It sounds like @onstartups read my post and implemented some changes! LOL! Which is good, I think that follower to following ratios should count toward their ranking system, which it clearly didn’t before.
BUT, thats not to say that what I had discovered previously with the update ratios isn’t still part of the algo..
If you take a look at martinbowling, with the most updates @ 10,862, he is up on followers & following less people, but only saw a grade jump of .9. Martinbowling was also a sufferer from the algo update, he was previously in the 99 point range until the update. (Sorry Buddy!) Compare him to oilman, who saw a similar rise in updates and followers, but is following more people, whereas martinbowling is following less, and oilman saw more than a 1 point gain in ranking. (FYI — before the algo update, oilman and martinbowling were .1 apart in score)
The only evidence I can take away from this, is that the update ratio still plays, although less significant than before. And now we also see the incorporation of a following to follower ratio.
I see this as good progress for twitter grader. My recommendation for the next algo update? Figure out a way to incorporate user interaction (@‘s, retweets, replys) After all, it’s all about the conversation
People mentioned in this post: