White papers or other documentation of existing curation-rewards schema for Hive and LEO?

in HiveDevs3 years ago

After a very informative discussion about #hive-engine with @aggroed a few days ago, I have decided to pursue setting up a new ‘tribe’ (i.e. front-end and token) geared primarily toward #university-students. I am toying with different approaches to distributing #curation-rewards for the new token. My understanding is that #Hive uses some sort of time-based curation incentive that has something to do with the 5-minute-mark after posting (creating a sort of ‘race’ to identify valuable content), whereas #LeoFinance has switched to a model that does not include time-based incentives.

Can someone point me to white papers or some other clear descriptions of how the current curation-rewards systems function, for Hive and Leo (I want to evaluate both)?

Also, if someone has some already-available sample data for a handful of posts, showing time-stamped upvote data and, if readily available, time-stamped reply data, that would be awesome (so I can do some what-if analyses looking at different reward protocols).

Thanks in advance!

Sort:  

There is no whitepaper on it, but I can explain it.

Hive has what is called a reverse auction in addition to time based curation. Also note curation rewards are 50% of the post rewards.

The earlier you vote, the more curation rewards you earn as you can only earn rewards on those who vote after you (and an amount of your specific vote). This encourages people to vote as early as possible to get the most rewards. To prevent people from blindly voting on content just to earn the most rewards, the reverse auction penalizes anyone who votes within the first 5 minutes with a sliding scale. This system encourages people to vote on the most popular content as early as possible but not too early that you forfeit all your rewards.

If you vote at 0 minutes, you forfeit 100% of your curation rewards. If you vote at 5 minutes, you forfeit 0%. 2.5 minutes you forfeit 50%. It is a linear reduction from 0 minutes to 5 minutes. (Historically it used to be 30 minutes, but then was changed to 15 minutes, then 5 minutes). Future hard fork will likely change it to 2 hours and remove remove any benefit to being "first" in the first two hours, in other words anyone who votes within 2 hours will get the same curation reward based on their stake (linear rewards, which explain in a second). After 2 hours, will revert back to normal time based curation rewards.

There is also a curve on Hive, in short this means posts with higher rewards will snowball and get even more rewards and low reward posts will have a more difficult time getting more rewards until it hits a certain threshold (around $10 they intercept). This curve is to discourage spamming for low rewards as it is difficult to find and counter this abuse. Some tribes implement a curve on post rewards, some do not.

Leo (and my tribe STEMGeeks) uses a linear distribution model, where you earn 50% of your vote regardless of when you vote. This encourages people to vote for what is "good" rather than what is profitable, and tries to encourage manual voting. It does not however take in account the lazy factor, and people just sticky votes as it no longer matters where you put them. The simple fact is, manual voting is a full time job and can take hours a day.

Both systems have their benefits and their problems.

Thanks for the clear and concise explanation!

I am contemplating a system that rewards auto-vote curation differently from manual votes. Within the existing framework, is there a straightforward way to 'flag' a vote as manual vs. auto?

No there is not. You can only build systems that encourage certain behaviors but there are limitations with a tribe.

Is there anything within Layer 1 that would be able to readily identify a vote as having come from a bot vs. being manually cast?

In other words, would it be straightforward for Layer 1 to add such a flag, or would that require extensive Layer 1 changes? My guess is that it would require extensive changes, but just 'wondering out loud' here.

There is no metadata on votes that would distinguish between a manual vote and an automated vote. There have been talks about adding metadata to votes (i.e. "app" flag) but no progress has been made on it and it won't be in the next hard fork.

That being said, as a large stakeholder, curating "manually" is a 4-12 hour/day job. It's hard to expect everyone is able to do such a thing.

I will try to post an explanation of what I am trying to achieve sometime soon.

The essence of it is to:

  • encourage quality manual curation (both for large and small stakeholders)
  • not severely penalize large stakeholders who don't have time to manually curate
  • encourage autovoting to accentuate quality manual curation rather than overwhelm it, stifle it, bury it, or manipulate it (intentionally or unintentionally)

I plan to detail a proposed framework, with adjustable parameters, that should be able to accomplish the above objectives. Granted there will be tradeoffs (hence the need for adjustable parameters).

The framework draws from the field of manufacturing process control.

you can do an onsite distribution, i don't know how to make it transparent, but as a site owner you can see most relevant content and reward it independent from votes, for example a share of ad revenue.

Is like the miner, but a bit different :)

How come it takes so much time to manually curate as a large stakeholder?

Sorry, I'm new to this and I'm still learning my way around the more technical stuff

Posted Using LeoFinance Beta

Because I don’t believe dropping $30+ upvotes 10 times a day. I prefer to spread a more realistic $2-5 votes to more people. This means even more time spent curating. It is a lot of work to find 50-100 unique quality posts a day to vote on.

This is probably the best explanation on this topic I've read so far. As a newcomer, it was a very valuable read! Thank you!

Posted Using LeoFinance Beta

If you have questions or need help with getting a tribe setup, feel free to hit me up on OpenHive.chat, Discord, Keybase, or BeeChat. The most critical thing is getting your tokenomics right as you can't change them later and it can drastically affect the success of your tribe.

Thanks! Will do.

A short example of what some of the voting and timeframes look like:

image.png

I voted on it after 25 minutes and then a really large account came in hours later and voted on it strongly and I got some really good rewards from my "curation" first.

image.png

This information can all be found at HiveStats.io which is a great site created by the @leofinance team. Valuable information in there!

#Hive uses some sort of time-based curation incentive that has something to do with the 5-minute-mark after posting

I think, its changing to 1 hour - may be @blocktrades can point you best.

The 5 min window was designed to reward the curators who find the future popular posts.

Like I find this post, vote early on and if it gets popular, I as a curator get rewarded for it.

This migh work wit large numbers of users but as it is now, the popular authors get recognized and in time curators vote for the person, not the post. Automation is also a big part of this.

For small groupl, linear curation as it is called might be better... no mattter the time voted on, or how many users vote after you. Its is also much more simple to understand. The curator has the freedom to vote whenever he likes.

In general curation has been long debated here .... how to gemify it, but as it stands now, simple and no gamification might be better option.... or maybe we just havent find the right way to gamefy it :)

I have posted about curation here
https://hive.blog/hive-167922/@dalz/is-gamifying-curation-rewards-necessary

Another post by the former official developer
https://hive.blog/steem/@vandeberg/reward-curve-deep-dive

Thanks for bringing this up! As a newcomer, I've always found it difficult to understand the curation system.

Posted Using LeoFinance Beta

Your doing gods work!!

Thank you so asking this as I have been wondering for a few weeks exactly what was the difference between now and the pre fork curation!

Welcome aboard.

The mechanics are much the same, but the math has been changed from this very dated post.
https://peakd.com/curation/@miniature-tiger/an-illustrated-guide-to-curation-from-the-simple-to-the-complex-with-real-examples-from-past-posts-part-1

If it was me, I would want to retain some aspects of 'find it first, get rewarded', especially with a young audience, but it has been a friction point for hive.

I've been waiting for Aggie to get really rolling, and here we go!