Curation reverse auction changing from 15 minutes to 1 minute in hard fork 21

in steem •  5 months ago 

Pull request

A change that was talked about previously was reducing the curation window further from 15 minutes. Hard Fork 20 reduced the window from 30 minutes to 15 minutes.

The current proposed change in Hard Fork 21 will reduce this window to 1 minute. The idea is to minimize the frustration of voting on a comment or post and it is too early.

If the Hard Fork is approved as is, the new curation reverse auction window will be 1 minute.

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!
Sort Order:  

This change favors bot voting, who the fuck proposed this? (excuse my french).

Not only bot voting but bots in general (often owned by whales).

  ·  5 months ago (edited)

The bot votes happen anyway. The idea is to get them over with in one minute and then people can vote without needing a stopwatch.

With the increase in curation rewards from 25% to 50% we expect that curation, including bot/autovotes, will get more competitive with more votes during the timer on known authors and more of those happening earlier in the timer. Which means that more of those curation rewards will return to the pool and result in higher payouts elsewhere. It also means that humans will be increasingly shut out by waiting or by trying to jump ahead.

Bots can play their games easily enough within one minute on the sure thing votes and everyone else benefits from both a more intuitive voting experience and more rewards returned to the pool and up for grabs elsewhere.

It does remove the cognitive load of trying to time the votes. Popular authors are setup for autovotes from big accounts anyway but with this change it will be difficult to front-run them. With the change to the rewards curve smaller accounts will have two options: either set earlier autovotes and forfeit on curation or vote on other not so popular posts and also lose some rewards with the non-linear portion of the new curve. Intuitively it seems to me that bot voting will have an advantage. I will give you guys the benefit of the doubt until we have enough data to analyze the change in behaviour that this adjustment will produce. Thanks for the explanation.

If people access their feeds often enough AND utilize notifications for their favorite authors it'll be a good change, otherwise there'll be other kinds of frustrations.

Thanks for getting the word out! Definitely this helps many cases where I hesitate to vote early. I do suspect that many current voting services do not have the granularity to vote exactly where they want though hmm..

It will be harder to beat the auto votes with a one minute window. Steemauto, for example, is only granular to the minute. I wonder if they'll change it to seconds.

Well, it is kinda stupid and pointless. It's easy enough to get a bot to wait. It's actually more frustrating when you read a new post or comment and have to come back in 15 minutes to upvote it. Sometimes you'll even forget.

It's a bit disconcerting how many major changes are happening at once though. Some of these should have been put into their own hard forks and given time to determine their effect.

So, HF21 is out for testing and there are still features to be implemented?
Shouldn't there be a code freeze?

It was frozen, but this was a hot topic before that was missed on the final chopping block. Because it is a simple three line change, it was added. Each additional hard fork introduces risk so on one hand, you want to do as much as you can each hard fork, on the other hand, you don't want to change too much so it becomes a balancing act.

I don't remember this being discussed here anywhere.

It my be only a 3 line change, but this can have huge impact.

Anyway, no one did any end user testing yet, AFAIK, so it shouldn't matter

Continuous improvement, I guess... development should not stop because a specific version of the code is being tested. But I am not sure how much time it’s agreed as safe here.

Posted using Partiko iOS

Even with ci, there are certain delivery dates, with approved targets, features, etc.
But this seems just random.

  ·  5 months ago (edited)

Then I would prefer to delay the HF then having a bad experience, like on the last one. If that needs to get into the HF.

This isn't really a code change it is just changing a constant. Due to some other bug fixes that were needed there will be a respin on the release candidate offering an opportunity to get this in. It had been previously discussed but the ball got dropped along the way.

AZ-5 is also just a button...

Let's hope for nothing like that Ж)

So if i watch a video, or read a five minute read, or listen to a song.... I have to vote, first?

Could somebody please think about this.

If you want to be competitive for curation rewards, yes that will be the case.

  ·  4 months ago (edited)

Only if you are confident the video or five minute read will actually by voteworthy sight unseen. If not then you may vote first only to find that those free downvotes end up turning your blind vote into a waste, or at least low return.

If you are confident it is votewothy without examining it first, others will be too and much of the curation rewards will be returned to the pool and used elsewhere, which is good; this is not valuable curation.

Interesting. This pushes the competition for curation rewards into a period between 21 seconds (block 7 after the post - equivalent of 5 mins currently) and 48 seconds (block 16 after the post - equivalent of 12 mins currently). A total of ten blocks.

That could mean a lot of votes in each block for popular authors where curation rewards are high.

Out of interest, how many votes can fit into one Steem block?

Also, what happens if someone sends through a huge post, very close to the block limit, at 30 seconds (or a number of smaller posts of an equivalent size)? Would the votes push into the next block?

Asking for a friend.

  ·  4 months ago (edited)

Votes are 100 bytes more or less (depends on the size of the permlink), so theoreticaly hundreds can fit in block (64 kbytes). I doubt this will be a major constraint, as very little content even gets that many votes total.

Also, is there a mathematical reason it doesn't make sense to vote earlier than 5 minutes, or is that simply a property of system conditions? Whether the linear penalty curve is really ideal is something I've been meaning to look at and may become more important if the window is shorter.

Also, is there a mathematical reason it doesn't make sense to vote earlier than 5 minutes, or is that simply a property of system conditions? Whether the linear penalty curve is really ideal is something I've been meaning to look at and may become more important if the window is shorter.

I think it's because you lose rewards linearly as you get earlier in the reverse auction period, but the post payout has to grow at n-squared to regain them. So in a simplified example under which you are voting first:

  • $1 vote at 15 minutes would require $16 post payout for a curation reward of 100% of vote value
  • $1 vote at 7.5 minutes would require $64 post payout ...
  • $1 vote at 3.25 minutes would require $256 post payout

The post payout becomes the limiting factor.

Votes are 100 bytes more or less (depends on the size of the permlink), so theoreticaly hundreds can fit in block (64 kbytes). I doubt this will be a major constraint, as very little content even gets that many votes total.

Is there a potential attack vector through which a group of accounts spams the blockchain with very large posts through blocks 7-16 (counting from the popular post publication) blocking anyone else from voting? They could tuck a vote on the end of their spam to grab the curation rewards at block 16?

Currently it would be hard work, since there's 150 blocks in that key voting period, but if you take that down to 10 blocks it seems more feasible. Even if you only hit 5 of them (equivalent to 3.5 mins delay in voting currently) there would be a fair increase in curation rewards.

Thank you for pointing out the potential attack vector. After bringing it to the attention of the devs, they have decided to back off a bit and recommend 5 minutes for now instead of 1 minute. In the future there may be some better protections against flooding attacks making it safer to reduce the timer further.

Cool! Glad to have been of use!

We typically have around 20 votes in a block, but we could do a lot more than that.

If two votes are in the same block, they act as if they voted at the same time regardless of where they were in the three seconds.

Anything that doesn't fit in a block will go in the next one but as far as I have seen we don't typically fill the blocks.

So in other words, after Hf 21 ... start curating the top people at 1 minute 🤦🏼‍♂️🤷🏼‍♂️🤣

Posted using Partiko iOS

I suspect that will happen a lot, but it will be a race to zero as many people will be doing it and trying to one-up each other by going earlier and earlier. This will result in more curation rewards being returned to the reward pool or people moving off to find authors who are not as competitive they can push above the curve and get good rewards. Some will vote as they always have, just earlier.

when does this go into effect?

When and if Hard Fork 21 gets approved. There is no execution date at this point.

thanks !

Well, I guess it was a bit annoying having to wait 15 minutes to upvote a comment and regarding people just voting posts for maximum profits, its really just the same when it can be done automatically with a bot at the right time so in my opinion there's no difference between a vote at the 1st minute and on the 15th minute.

its really just the same when it can be done automatically with a bot at the right time

I don't think the bots are so accurate, are they? I don't really know as I have never used them.

I think no one will upvote comments much after hf21 with the new curve.

Posted using Partiko iOS

Not necessarily. Look at the nether region of the new pseudo-linear curve and a small upvote could be worth just half what it would be under linear, yet the upvoter gets 50% curation rewards on it - comes to the same 25% return... just the author gets a lot less, like one third (ouch!)

I think the wait time is counter productive. Those who know it and how it works wait to vote etc. Those who don't know it most likely are giving up rewards and have no idea they are doing so. Keeping the reward the same throughout the period of time makes sense to me.

There should be less confusion and a simple system that is friendly to new community member of steemit. Again 50/50 rewards makes sense. No waiting to try and get an advantage u are simply rewarded equally.

Posted using Partiko Android

Erm... Bots will farm all the curation rewards...

Pardon my slow brain right now, though I could use a translation into an easier to understand clarification:

Does this mean that while it used to originally bethatd we’d want to vote only after 20 min after a post to receive curation rewards (or we wouldn’t get any if voted earlier), then 15 minutes...

This now means we can start curating one minute after a post is published and receive rewards - and we’d want to be the earliest to vote to receive the maximum share of the curation split?

Please confirm or correct if I’m mistaken, thanks... 🙏

Before Hard Fork 20 there was a 30 minutes reverse auction for curation rewards. If you voted immediately, 100% of the rewards would go to the author. If you voted at 15 minutes, it would be split between you and author 50/50. At 30 minutes you would receive 100% of the curation rewards.

Hard Fork 20 changed two things, it reduced the window to 15 minutes and it returned the rewards to the reward pool instead of giving them to the author.

With Hard Fork 21 (as it is written today) the window will be 1 minute.

Thanks to @jga for leading me here. Have recently analysed both the 50% curation rewards and this 1 minute reverse-auction, as I just got to know about the latter - not that it was paraded much in all the EIP promo posts.

The new curation rewards have an unexpected (or perhaps designed) twist to them in that passive income programs such as vote-selling in whatever form will earn less in absolute terms but actually more in percentage terms when compared to the author-rewards ;-) This should be more widely known.

This affects this reduction to a mere 1 minute because otherwise combining the drastic reduction in low value votes as envisaged in the pseudo-linear curve plus the reverse-auction curation losses would mean that many curation-rewards-seeking votes would make very little, even with the doubling to 50%.

What the EIP actually does is make it harder for smaller accounts to earn above average curation rewards but make it much easier for large upvoters to do so.

I don't see anything to make human-curation more profitable as those votes come much later, apart from those who work within a curation group that can then deliver large upvotes. But the pitch that EIP would make bot-curations less profitable than human-curations is just not true - what it does is make large-vote-curations far more profitable in steem terms, bot-curations more profitable in percentage terms and human-curations a little bit better off.

I don't expect any changes, just to inform users what is most likely to happen and to adjust expectations accordingly.

good project, which will help good people of this house. happy weekends!

Posted using Partiko Android

This favors whales, bots, insiders, staff, witnesses.
On the backs of the content-creators.

That's one of the changes I like in the upcoming hard fork :)
Can't be bitchin about it all the time, right?

Posted using Partiko Android

Can’t say I’m totally clear on who benefits from a one minute window. If early voting is incentivized, I don’t see how the whole system doesn’t become steemauto on the same list of creators+bots ruling exposure.

Speaking of unsustainable systems, my own posts are now getting downvoted by a complete stranger just for interacting in the comments of some of the palnet folks. Good times.

Awesome. Hate having to worry about the time since posting before I apply a vote.

A general question about curation: do votes on comments behave the same as votes on top level posts? By that I mean is there a curation window and do early votes garner relatively more curation rewards than later votes?

The blockchain sees no difference between comments and posts.


Thank you for this information

Now they tell us?

I just hope hf21 doesn't take as long to 'fix' as it did to fix hf16.

I don't like this change, it makes bot / automated votes > users organic votes. No user can compete with bots and put vote early enough to get good curation reward.

The idea is to minimize the frustration of voting on a comment or post and it is too early.

Currently, the best reward is when we vote, say, between 5 and 10 minutes. I don't think it is currently a big deal with voting too early then.

  ·  5 months ago (edited)

Currently, the best reward is when we vote, say, between 5 and 10 minutes. I don't think it is currently a big deal with voting too early then.

It used to be 12 minutes...then 10, then 8...

Edit: also depends on the size of vote

Yeah, I know. It depends on sum of rshares before our vote, sum of rshares of our vote, time of vote and total sum of rshares after payout. But for most users in most cases the best time to vote is between 5-10. And the smaller the vote the more profitable is to vote earlier. On the other hand whales should vote at the end of the curation window.

Some time ago I made a analysis about curation, but haven't post it. Maybe it's time to do it now. There was a lot charts as below.

I revoted this comment because of the fool downvoting it ;D

Yes, the curation-reverse-auction period is an interesting (was an interesting) bit of complexity within an otherwise naive economic code. I've just written in the last couple of days on the effect of 50% curation and this new 1 min period. We need a few people who understand the maths! ;-)

The original idea set at 30 mins was precisely to stop bots voting super-early. My analysis suggests this timing change was smuggled in purely to offset the steep loss in upvote values for low votes.

It's all curation-reward-mining - which is not so bad if everyone did it.

So if I’m using Steemvoter and have a higher than average voting power, 10 min would be ideal? 🤔

Not completely sure. depends on the weighting etc and when other people vote before you - I generally go with the "vote whenever I vote" paradigm :D

It's complicated ;-) and unstable. There is no perfect time, merely a moving target that averages to better than standing still.

I think this is a good chance, at least from my perspective of upvoting posts whenever I see them, and not trying to maximize my curation rewards. There are so many times where I have wanted to upvote comments, but decided to wait until they hit the 15 minute mark, followed by obviously completely forgetting to upvote them once the 15min mark is hit.

not trying to maximize my curation rewards

decided to wait until they hit the 15 minute mark


The thing is that the author also get nothing if I upvote before this point. So there is not even a point to upvoting anything before this mark.

Of course he does, the author reward has no time window :)

Well, it's not nothing, but it's also less than I can give if I wait until the full 15m mark. If I upvote a comment after about a minute, my small $0.1 vote will be practically worthless.

It does not work like this. Only curation reward is affected by voting before 15m. Not author reward.

This is correct

Damn, my mind is blown. I can't believe that I have misunderstood this for such a long time. Thanks for the clarification!