RE: [Steem] Steem Per Mvests Leads to Bot Island
Thank you for the contribution. It has been approved.
Hi @lextenebris, glad to see you on utopian as well, great start! Sorry for the filters left out "for simplicity" :D But you worked out the details exceptionally well!
What you found with deposit and withdraw account being different are withdraw vesting routes. Those are not available via steemit.com, but for example with the steem cli_wallet or steem-python/steem-js. If you power down on steemit.com you automatically get the appropriate amount of STEEM into your own account over the course of 13 weeks. Under the hood, this process is much more flexible:
You can withdraw parts or all of the powering down SP to one or more other accounts controlled via percentages. You can also choose if the powered-down SP should end up as STEEM or as SP on the target accounts. The latter allows you to move SP from one account to another without going via STEEM. What you mostly saw in the data are 100% withdraw vesting routes where the full amount of SP goes as STEEM to another account. This way an account can be powered down to really 0 SP without locking the funds with 0 bandwidth. Withdraw vesting routes result in two fill_withraw_vesting ops in the blockchain per week, one for the powered-down account and one for the receiver. In the case of a 100% vesting vithdraw route there's one operation where the powered-down account withdraws X Vests into 0 STEEM and one where the receiving account withdraws 0 Vests into Y STEEM (IIRC?). Lower percentages than 100% seem to give lower steem-per-mvest values in this case. These are the "outliers" in my graph.
Since withdraw vesting routes are comparably rare, the median value like you chose should give correct results for the steem-per-mvest. Maybe the daily/hourly/... upper limit might be a a good candidate as well. Or simply taking only ops where deposit and withdraw account are the same...
A lot of those powered-down "clusters" are likely mined accounts from the early Steen days, where the mined SP is transfered to the main miner account. The powered down accounts related to sigmajin are no unknowns, I came across them before because they all share the same Steem keys and were used to flag zappl posts in the past. @abh12345 also found them recently analyzing the flagging behavior in the last couple of months. With the same active key for all accounts it's easy to power them down in the same transaction.
You can contact us on Discord.
[utopian-moderator]
Not so much a great start as it is a great return to form. Even if Utopian wasn't exactly keen on my 3D design work for an established project. Still, flowing liquid under the bridge.
I was really hoping that the "filters" were going to be a lot more interesting in terms of how they simply cut out stuff that was excessive, but it turned out that whatever lived in there was exactly the stuff that would attract my attention.
It's just like having a sore tooth. I am the tongue that cannot resist poking it.
Right, which is largely what I was the arising – but part of this system still eludes me: what triggers the creation of this operation? Three quarters of the time, the from and to account are the same account, which makes a lot more sense for most purposes. About 1/3 of the transactions that take place in the last year were powering down accounts shifting the results to another account – and not just one or two accounts, but entire constellations of accounts, draining into a more centralized one. As I have tinkered with making the time horizon closer, I've found a strange thing! The percentage of the
fill_withdraw_vestingoperations weren't dropping as rapidly as we were looking at an ever smaller cross-section to see where the activity was coming from.I even shrunk the time horizon to one day, and observed what the rest of the layout ended up looking like, the amount of second party investment transfers were much closer to those transactions than 1/3.
One could say with some degree of surety that whatever is going on under the hood causing these events, so I have to roll back to ask about the initiatory choice that causes this particular operation to get signaled? Does it apply whenever an account is set to power down?
Like I said, even when I reduced my temporal horizon to 30 days, even a week, I still had far more of these vesting route operations than made sense. To me, that suggests that they are all actively engaged in sending that vestment to the central dumping spot and collecting it there, until it can be dumped for fair amount.
I'm not sure that I am ready to buy the "old clusters" rationale wholesale, at this point. After all, three or more of the highest rated, highest trafficked accounts on the blockchain are still listed in this list.
At some point, it would be very useful for us to lay down the comparisons and say which are possibly better at solving the problem.
In the meantime, I suppose I've discovered enough insanity floating around the world for one night.
Exactly! The power down is started with a
withdraw_from_vestingoperation. Seven days after this comes the firstfill_vesting_withdrawoperation that converts 1/13 of the VESTs into STEEM based on the current steem-per-mvest ratio. 12 morefill_vesting_withdrawoperations follow in the remaining 12 weeks so that after 13 weeks all VESTs from the initialwithdraw_from_vestingoperation are converted into STEEM.That's – interesting.
It's particularly interesting because even as I shrunk the time horizons on my query, these operations decreased in total volume, both standard and nonstandard, about as you would expect – but the transactions like this, with a different from and to account, became a larger proportion of those transactions in general.
Which leaves us with the question of what these things actually represent. I am willing to accept that some of them, particularly the smaller clusters, just represent a small bot host owner powering down their bots which gained SP one way or another. That's a perfectly normal sort of activity, even if I'm not particularly on board with how useful bots are compared to how abusive they can be.
It's the ridiculously large clusters, with dozens to hundreds of nodes, that concern me. They have enough accounts in them that they could probably keep a number of them rolling over every week and generate a consistent payoff almost daily assuming that they could continue to farm the SP somehow.
An extremely small number of accounts is responsible for 1/3 (or more, depending on how we split it) of the this particular type of transaction of transactions which occurs on the blockchain every day. That seems concerning to me.
I suppose the next step is to try and figure out what these things are doing, if anyone cares.