You are viewing a single comment's thread from:

RE: Witness consensus status to fix the actual steem’s economic flows (ENG)

in #witness-category4 years ago (edited)

I'm only a low-five-figures investor but I will be cashing out and seeking opportunities elsewhere if 50% curation is the consensus. We don't have enough incentives for posting content now, which is incredibly obvious if you look at the posting quality and quantity after the fork. Further reducing it basically makes Steem a desert, and you can fight over your curation all you want.

I do appreciate the general disapproval for superlinear, though.

Sort:  

Curation plays a key role in rewarding the good content. Otherwise we have a reward pool of X STEEM/day spread out over all kinds of garbage and abuse and the valuable content contributors get 75% of very little.

The goal of these changes is absolutely to get better voting and curation which means more rewards flowing to those who deserve them.

I respect that you may disagree on this but I personally find such strong disagreement a bit peculiar. From 75% to 50% there is a 33% reduction in rewards. I don't know how you can be so confident that better voting and less voting motivated by content-agnostic extraction from the pool won't shift enough back to the good content to make up for this 33%.

Curation plays a key role in rewarding the good content.

No it doesn't. The only way to believe this is if you're completely bamboozled by the white paper. In practice curation doesn't do this at all. All it does is reward predicting what other people will vote on.

I don't know how you can be so confident that better voting and less voting motivated by content-agnostic extraction from the pool won't shift enough back to the good content to make up for this 33%.

The current optimal curation strategy is to cast an extremely broad net of tiny votes in order to catch the occasional "hit" of rewarded content and win a small amount of SP each time through the n^2 curation algorithm. Look at how well @cheneats does as an example. That's the behavior you're encouraging here.

No it doesn't. The only way to believe this is if you're completely bamboozled by the white paper. In practice curation doesn't do this at all. All it does is reward predicting what other people will vote on.

I'm not bamboozled by the white paper. I've explained (for years, often a lone voice on this) how downvoting is essential to anchoring the payout values (and therefore curation) to actual opinions on value of content/contribution. The white paper and the whole notion of downvoting only as an anti-abuse tool is getting this completely wrong (though to be fair with n^2 rewards there is less difference between the two; the original superlinear model with cheaper downvotes surely would have worked better as well).

The current optimal curation strategy is to cast an extremely broad net of tiny votes in order to catch the occasional "hit" of rewarded content and win a small amount of SP each time through the n^2 curation algorithm

That provably isn't the optimal strategy. I can easily improve on this by identifying some posts that won't 'hit' and not spreading to them. But to some extent this is true today because of insufficient downvoting. Literally any garbage can end up being a hit if it gets bidbot votes, and then it generally isn't downvoted enough (i.e. down to zero) because the cost of doing so is too high. Still, I think you can do better.

A large number of tiny votes on decent-to-good content is a useful form of both curation and content discovery in that it separates the possible wheat from the chaff. For example, I often upvote my own comments in large threads by 1% not to earn curation rewards or the few cents of self-voting-reward but to separate them from the huge sea of (often low value) comments.

The second level of curators then has less to look at, and can begin to shift rewards among the identified 'not worthless' candidates. This is good division of labor and value add all around, especially if you consider a scaled up system with much more content added daily.

The second level of curators then has less to look at, and can begin to shift rewards among the identified 'not worthless' candidates. This is good division of labor and value add all around, especially if you consider a scaled up system with much more content added daily.

In order for any of that to work, the large majority of the stake has to be voted naively, so that the people doing the algorithmic discovery have a large pool of content-based votes to be rewarded from. And you get naive voting by reducing the incentive to optimize, not the reverse.

If you have a first-level of curators that does basic discovery, and a larger (stakewise) second-level of curators that does higher-value discovery, you still need an even larger third level of "uncurators" who just want to vote on cool shit and don't care what their rewards are.

That remains the key population for making Steem work.

In order for any of that to work, the large majority of the stake has to be voted naively, so that the people doing the algorithmic discovery have a large pool of content-based votes to be rewarded from. And you get naive voting by reducing the incentive to optimize, not the reverse

It works if overrewarded content is downvoted. Curators can pile on (or perhaps 'reverse pile on' is more accurate since it is anticipatory) because A expects B to curate/vote and B expects C, etc. But if D (along with E and/or F) comes along at the end and declares this to be trash and downvotes it to zero then A, B, and C all wasted their vote power (a costly mistake).

Even without any preference-motivated upvoters (and there will always be some) this can still work. As a curator you have to upvote something or your vote power is wasted. If you spread your votes thin or even vote randomly, this is superior to not voting. Once you do that, then choosing based on what you most expect to not get downvoted is superior to voting blindly. Simple model of course, but not absurdly off.

Somebody has to be the last voter who loses in this scenario. If not a naive voter, who's getting what they want on a personal-social level, then it's a curator who's not doing well enough to be successful. Over time the unsuccessful curators will drop off and somebody else will become the unrewarded top of the pile, until it has recursively reduced the curation-optimizing stake to the level the naive voters can support.

I don't think you have refuted my logic even there are NO naive voters other than downvoters (who count as a form of naive voter in the sense they are unrewarded).

Yes, there will be some equilibrium of curators. That's okay.

I don't understand your model here. Is it that you expect there to be a significant population of people who both desire curation rewards and are satisfied to lose at it? Do you think curation at 50% is so great that even the worst curators will be better off doing it than anything else with their money/time? (There might actually be enough inactive Steem to make that true in the immediate sense, but I'm doubtful it's sustainable. Though I really have no way to model future activity %s. I guess free downvotes help this too, and I'm going to have to think through that part some more.)

If the small curators are being paid by the medium curators and the medium curators are being paid by the big curators, who's paying the big curators? It's still zero sum and somebody has to lose.

Except that everybody in that system is getting paid by inactives and people who voted on things that got flagged, which is still zero-sum as a whole, but I need to think about it some more. There's a weirdness there of creating a system that sort of depends on the existence of abuse that I kind of like, but is super-weird in a game theory sense.

I think with 50% curation rewards we are going to see more bots like @nfc and people like me who would power down to lease SP (don't have money to invest), then follow the trail.

It too expensive to produce content but cheaper to get similar rewards from curation. Or at least thats what I am thinking.

I was following the NFC trail with around a 2x vote for a while last week to see how it did. (Steemauto won't actually do that so I kind of faked it.) It was maybe a little more profitable then the cost of leasing, but not a lot, and of course there's the risk of it losing value in an emergent market, since you'd have to commit to the lease for a long term to get a good rate.

I think either delegating to NFC directly or building your own curation bot are better options right now. @night-heron's doing reasonably well putting tiny votes on your comments. :)

Here is another @nfc experimental post. I will obviously think of coding a bot myself if @nfc doesn't work.

BTW I saw @night-heron was voting my comments and got good curation. You'd be better off with 50/50 system. ;)

I don't doubt that The Mespotamians would have a substantially larger return under 50%. But the cost is it making our mission far, far more difficult.

You're looking at it too statically. What if 50% curation more than doubled the upvotes you got?

Posted using Partiko Android

What if witches flew down and gave me a winning Powerball ticket?

Why wouldn't large accounts spread more of their voting power to other's posts if curation was 50% and not 25%?

Posted using Partiko Android

Because people who are interested in passive income are interested in passive income. No one's going to magically become a participant because you change the number around.

Also you're the ones trying to ruin user acquisition, the burden of proof is on you.

There will still be passive income. Instead of delegating to a content-agnostic voting service you delegate to a decent curator and share the rewards.

Just as easy access to vote bots change the dynamics of how people act, easy access to highly profitable curation programs will also change how people act. Investors will go where the money is. Currently the incentives in place encourage voting with bots, regardless of the content quality. Incentivizing curation incentivizes valuing content that others will also value (i.e. "good" content).

No, it also incentivizes voting with bots, just @nfc type bots. Even the curation increase with HF20 has done that. I've been playing around with some of my own. It's not about content quality, it's just about vote-racing, mostly in front of who Steemit is voting by proxy through the @misterdelegation recipients. Which makes sense, since curation has never effectively promoted content discovery.

Also, it's pretty clear that the peak of "quality content" on Steem is content that is perceived to promote the Steem price. If you reduce the diversity of content by concentrating votes you're just hastening the Bitconnect-ization of Steem.

It's not about content quality

The only thing in this system that has ever been about content quality is downvoting. That's why reducing the cost of downvoting is essential to getting the rewards where they do the most good.

This is a package of changes and looking at them piecemeal is misleading or in some cases deliberate spin:

  1. Increasing curation shifts the 'go where the money is' incentives toward curation rather than self-voting. This would produce more curation but would not work to produce good curation without #2.
  2. Cheaper downvoting means more upvoted garbarge will get downvoted, and therefore curation (see above) needs to focus on non-garbage to get a good return. This would not work to produce good curation without #1 (actually I personally think it would work, at least better than the status quo, but it works better along with #1).

Looking at the pieces separately doesn't show the whole picture.

This is all focusing on abuse rather than usability, which is one of the Steem development process' biggest flaws. Yes, fewer votes might go to crap, and if that's your goal, bravo.

But the cost is fewer votes going to new posters. Something that we didn't start with enough of and have been dropping like flies.

Our argument here seems to be about whether promoting user growth is more important than reducing abuse. I strongly believe that to be true.

since curation has never effectively promoted content discovery.

And this is the problem we're trying to fix by incentivizing more curation so that it can effectively promote content discovery.

If we assume more people move to curation bots, would the quality go up, got down, or stay the same? Curation may not be perfect, but I do think it's better than what we have now.

"Quality content" is certainly skewed towards those with the most voting power and it makes sense they want to increase the value of their holdings. One important way to do that is to show the Steem blockchain via Steemit, Busy, Steempeak, Steemmonsters, etc working effectively. That includes demonstrating an effective curation system bringing good content to the surface. By that, I mean it's not just "Hurray STEEM!" articles that will, in the long run, have a positive impact on token price. It's also important (for those who understand and consider long-term value) to promote and demonstrate the effectiveness of the stated (and valuable to investor) goals of the Steem blockchain.

And this is the problem we're trying to fix by incentivizing more curation so that it can effectively promote content discovery.

"It doesn't work, so let's increase its influence."

If you could come up with a curation algorithm that actually worked for content discovery, there might be some argument for increasing rewards. But the current one demonstrably does not do that.

If we assume more people move to curation bots

Curation bots is too generic to reach any conclusion. It depends how those bots are programmed, what inputs they receive, etc.

We can reason based on economic incentives and those incentives apply equally to humans and bots (created by humans based on the incentives). We can say with near certainty that changing the incentives will change the behavior. It is more difficult to predict exactly how the behavior will change, but we can still make some educated guesses (and accept that despite doing so we will sometimes be wrong).

It's not just any number. Besides it's part of a set of proposed changes. Secondly, all I'm trying is to get to the bottom of this by talking to people having different opinions on this.

Posted using Partiko Android

It doesn't need to double. Reduction from 75% to 50% is only a 33% reduction.

Yeah... 50% won't get vote sellers the desire to stop selling their votes, it will just give them more money for not curating themselves and keep selling their votes to make money. 50/50 is not solution at all, it only incentivizes more vote selling anyways as I see it.

Vote selling pays about 100%. 75/25 or 50/50 doesn't change it or increase incentives at all, nor does it pay any more to vote sellers (about 100% either way). It increases the relative incentive to not sell. Not enough on its own but it narrows the gap significantly.

Coin Marketplace

STEEM 0.20
TRX 0.06
JST 0.028
BTC 23584.81
ETH 1662.25
USDT 1.00
SBD 2.66