Altogether my POI went up by 1.18% (to 6.61%), before the 50k coming back and with some harvesting.
Has anyone drawn any conclusions regarding POI? Any new theories?
The whitepaper should be out within the next few weeks. Until then I go with
PoI(v) = 0.5*PoS(v) + 0.5*PR
v being the vested amount and PoS(v) = 9/76*10^-9 * v (so the total "vested" XEM is not the full 8999999999).
PR is the undirected PageRank of the account. An edge between two accounts is established when a tx>=10k XEM has been made and is vested in the target account. Don't ask me which damping factor is used.
There don't seem to be multiple edges between accounts, so sending large amounts of XEM is useless other than the 10k XEM being vested sooner.
There might be some cluster detection in place (there are various well known graph algorithms to do that), but at this point I believe this is not the case. Probably it will be implemented at some later stage or earlier if needed.
That being said, these are just my speculations. Don't trust me too much.
Edit 15.05.:
Turns out I was quite far from the truth. Well, it is a variant of PageRank at least, but involving cluster detection and weighting decaying net tx. Far more elaborated than expected. I take my hat off to the devs.
Have to do some investigating however why my tests looked quite like what I described above.
The simulations seem to suggest that splitting an account in 20 parts, waiting for them to be sufficently vested and transfering everything back to some other account seems to be the best way to gain
some importance score. Have to test that some day ::)