In brief, to downvote self voted posts and comments of those accounts with the highest self voting return on investment. That is, the users who use their voting power to direct rewards vastly back to themselves.
There's a much more in depth discussion of the rationale in the first post.
In version 2 of the algorithm, launched for a few weeks almost 2 months ago, there was a hidden systematic bug. It turns out that changes in delegations make using SP as a metric variable impossible (without running a node). At first I thought it was possible to record delegation changes over time and them work backwards to previous SP, but you would need entire knowledge of the account because of the way the blockchain records _un_delegation. When you undelegate it simply records a delegation of zero, but this makes reversing it difficult because you must scan backwards to find the original delegation value. This is an issue with Steem Reports too, though it usually doesn't matter too much. Be aware of this when comparing accounts with that service.
We made the decision to not run a node to replay the blocks over the period, instead went back to the drawing board to look for a better, cheaper solution.
Apologies to current and previous supporters for a lack of communication on this front. It was a case of one more thing ... one more thing! I thought I had it solved, pushed many patched, but eventually reached the dead end.
Instead of a payout value based metric, we have switched to a voting behavior metric. This achieves the same thing which a high degree of accuracy, but using relative values and some smart assumptions. We look primarily at voting frequency as a function of voting power regeneration to observe who is self voting most optimally. Scores produces are heavily mitigated by outgoing votes, so those who vote outwards in general will have very low self voting ROI score. Because it's a function of voting power generation and voting frequency we can take into account distribution of votes over time without needing to know the exact VP, or even SP of the account. Thus the scores are already relative to SP.
I've been testing this for a few weeks and working out the kinks. It has been live for about a week now and I intend to keep it running with some planned tweaks and keeping an eye on it.
We have already had some renewed support for the bot in terms of delegations to 
If you delegate to
Many delegators consider a delegation to be a way to actually improve the standing of their investment. Talk to some of them about it if you want to know more.
You can always check out what
There are always people complaining about this project, and how we "steal" from them, why can't we leave them alone. I would say it's not hard to get out of the radar for
The project is being expanded from self votes to a primary voter list to try to encompass tight voting cliques. After an initial period of research this will be trialed.
Everybody makes mistakes, and this project is no different. While every effort is made to analyse the blockchain accurately, if you feel you have been downvoted in error then comment on this post with details and I will review it. We are committed to only proceeding if things are done well.
