Superlinear

The goal of this post is simply to cement just one aspect of the recent Curation Proposal Fad in my mind so that I understand its effects better.

First, what is this n^x curve people keep spouting on about?

This refers to how much weight your vote has. For example, in n^2 weighting, someone voting with twice the SP (assuming full weights for simplicity) has 4x the weight.

But what does weight mean? The reward pool is distributed according to what fraction your weight has relative to the total weight. In a world with just a 10 SP voter and a 20 SP voter, that means the 10 SP vote is assigning 100/500 of the pool and the 20 SP vote is assigning 400/500 of the pool.

So: Twice the SP, Four times the influence in n^2 weighting. Personally, I'm not even sure why there should be any multiplicative bonus. Linear just makes sense in principle and is a much simpler rule.

The theme behind this kind of change is that it encourages concentration. Less posts to be rewarded higher values, and less people effectively deciding what is rewarded. This theme is already a problematic one: do we really trust those with stake to pick the right posts? Much less police themselves (re: downvoting). In a linear world, we can see that some of those with stake aren't doing much good with it. In a superlinear world, it's worse.

Self-voting obviously is more attractive in a superlinear weighting versus linear, when considering this change in isolation. And if the top holder has that much more influence, it makes it that much harder to counter votes of high SP abuse. (Note: in the referenced proposal, this is likely why we may need both discounted downvote and preferably smaller exponents, but just want to make sure I understand it without those components first.)

There is one variant that seems worth mentioning briefly, though a little awkward: capping the max assignable weight on a post. This would mean that even if a large stakeholder separated into many accounts of maximum weight, the power becomes linear relative to other whales (can only stack max weight votes). This still makes it a whale's game, but makes it so the first place whale is not that much more powerful than the second place whale, or that whales have a chance to stack votes to bring down a larger whale.

I was playing with curation formulas on the side as well just now and did not find anything too different when comparing to their vote value. Could be missing something of course.

Superlinear makes sense if it wasn't tied to stake, as suggested in this thought experiment. Though there is a comment there about how when we had n^2 in the past there was less self voting. I don't really see how that's possible. Unless there were more whale policing? I don't know. Maybe someone can clear that up for me. Actually if anyone can really point to why n^2 was so good for us (if true) that might help too. I wasn't here for it.

If there's any misconceptions I've fallen into for superlinear, specifically, I would like to know. Though I'm anyway following the discussion on the other posts and this seems to be correct.

I'm not sure if I'll get to analyzing all three proposed effects from this post by @kevinwong but suffice it to say that my line of thinking now is matching that of @smooth at the moment.

And just food for thought here. People cite that incentives need to align, but what if I told you they already align, but you have to pay attention to the time scale? As in... the long term value. For example, what if token investors start actually paying attention to token allocation and aren't happy with it? Basically, why aren't we sinking already? There's so many different things happening right now, and it's really hard to say. And all of this is kind of useless anyway because you can be like "well steem used to be 5$, blame the behavior. See?" But that isn't productive because you cannot isolate this bear market from general cryptocurrency bearishness. Wouldn't it be great if we had a market that reflected accurately the sentiment?

And about content. Are we rewarding content, or are we rewarding networking? Engagement? Which is it? And which reward scheme is better if we want to reward the latter? Do we want a full, rich economic system? Or simply a faulty content appraiser? I'm not saying things are peachy either, but it helps to know what the real goal is.

Anyway I seem to have strayed off the intent of this post. I'll probably find my way into the main discussions once I understand things better. Maybe.

Posted using Partiko Android