You are viewing a single comment's thread from:
RE: Screw Malicious Flagging: 1 Flag Vs 130+ Votes
The simple solution is to change the reward weight to: votingPower * sqrt(SP) so that SP doesn't have a linear effect.
The simple solution is to change the reward weight to: votingPower * sqrt(SP) so that SP doesn't have a linear effect.
I would even go for a logarithmic weight.
That will not work. If SP has sqrt voting power then anyone can increase their power by about 41% by simply splitting their SP into two accounts, and increase it even more by splitting into more accounts. Only the most aggressive value maximizers (which overlaps somewhat, but not entirely, with abusers) would do this, leaving casual users with even less power.
Having three accounts and doing everything the same on all three accounts is a lot of extra work.
Also it would reduce your author rewards. You can curate three or more times, you can't author three or more times.
I think most people won't bother splitting accounts.
You're probably right, most people won't bother, but abusers trying to maximize their profits certainly will. Which means more voting influence and ultimately more rewards will go to those who game the system, and less to the typical or casual users who don't. There are already bot herders with thousands of accounts. It isn't hard to manage a few more.
BTW, splitting up SP does not hurt author rewards. SP has no influence on author rewards at all unless you self-vote, which can still be done easily by having all of the accounts vote for the posts.
abusers arent most people and for them the extra effort is nothing
I was thinking about this. The Sqrt needs to be applied to the total SP for a post and not individual votes. That is,
Sqrt(v1*sp1 + v2*sp2 + v3*sp3...)and notSqrt(v1*sp1) + Sqrt(v2*sp2) + Sqrt(v3*sp3)When it is applied to the total, the splitting of votes doesn't matter to the payout.
True, however this still rewards splitting votes across posts. How does that matter? This would give far more rewards to self voting, and now self-voters would post more times with smaller votes on each for a larger total reward. Not only does that increase overall reward for gaming the system (and therefore less for everyone else) but it would incentivize posting more spam, low value comments, etc.
That's a great point. But I beleive that can be solved too by the same method. When you calculate the reward for a user, you calculate the total steempower incident on that user across all posts, and then apply
sqrt()on it, instead of calculating for each post separately.I'm not sure what you mean by incident on that user across all posts, but if I understand correctly then the abusers can create many new accounts and have each one only post once (per period). There won't be anything to combine across posts with only one post per account. This is even worse than just spamming because all those new accounts impose overhead too.
There are other implementation issues with this such as the computaitonal overhead of aggregating, the fact that the posts aren't paid at the same time, etc. But those aren't worth getting into because any attempt to impose sublinearity when there is no mechanism to force users to keep their stake in a nice bundle where it can be counted (and therefore penalized) is pointless. And even if it were, It is also very questionable from the perspective of STEEM value whether it makes sense to introduce a strong incentive for anyone who already holds STEEM to sell it and reduce the incentive for anyone who already has some STEEM to buy more. That sounds like a recipe for disaster to me. The whole premise is looking at things too myopically from the point of view of voting 'fairness' (questionable to me as "one SP, one vote" is also 'fair' in a lot of ways) and losing sight of the fact that everything rests on top of investors attaching a high value to STEEM. If that isn't the case then voting means nothing because you will be voting on nothing but crumbs.