Steem Analytics: Earning more than 100% of vote value through curation and effect of 50/50 on this practice.

in #utopian-io2 years ago (edited)

SteemAnalyticsSolo.png


1. Introduction and Theory

Almost two years ago I posted some theory about how curation rewards work. This included the conditions necessary for a voter to earn 100% of their upvote as a curation reward, such that the curation reward they received was equal to the value of the vote they added to the post.

The theory included this nice little diagram and an explanation (which I have abridged):

Ruleof4.png

"Blue: The voter adds the first $1.00 to the post payout and at this point would receive all of the curation rewards of $0.25."

"Green: The post payout needs to increase to 4x the original vote in order for the curation reward paid to the original voter to double from 25% of upvote to 50% of upvote."

"Orange: The post payout needs to increase to 16x the original upvote in order for the curation reward paid to the original voter to quadruple from 25% of upvote to 100%. At this point the post payout is $16.00, the total curation reward is $4.00, and the weight share for the first voter is 1/4 giving a curation reward of $1.00."

The link is here (note that the post was made prior to the change from a 30 minute reverse auction window to the current 15 minutes) :
https://steemit.com/curation/@miniature-tiger/an-illustrated-guide-to-curation-from-the-simple-to-the-complex-with-real-examples-from-past-posts-part-2

So what proportion of votes can earn 100% of upvote value under 75% author / 25% curation?

The above example would suggest the maximum "capacity" to be 1/16, i.e. 6.25% of votes could earn 100% of upvote value.

Of course, this example is not particularly representative. Most posts will receive their first upvotes within the 15 minute reverse auction window. For posts that are expected to receive high rewards, competition for curation can drive the earliest significant votes down under 10 minutes, with smaller votes arriving earlier still.

For a $1 initial vote made half-way through the 15 minute reverse auction window, i.e. at the 7.5 minute mark, the overall post reward needs to increase to $64 in order for the original poster to receive 100% of their upvote as a curation reward. This can be broken down into:

  • x4 to recover the 50% of curation rewards sent back to the reward pool by voting half-way through the reverse auction window; and
  • x16 as before to increase from the 25% curation rate to 100%.

This example suggests the "capacity" would be 1/64, i.e. 1.6% of vote amounts receiving 100% of vote value as curation rewards.

As a final example, consider a $1 initial vote on a post / comment that receives no other votes, or that receives less than $16 of votes in total. In such cases 0% of votes receive 100% of vote value as curation rewards. There would be "capacity" here for someone to vote prior to the initial $1 vote and gain high curation rewards but in many cases this will not occur.

So what is the overall percentage of votes that earn 100% of vote value as curation rewards under 75%/25% author/curator? Somewhere between the 0% in our third example and the maximum 6.25% in the first example, but probably something less than the 1.6% given in the second example.

Before I get into the analysis, how do these theoretical numbers move under the 50%/50% author/curator reward structure proposed by the EIP?

  • Example 1: Under 50/50 a $4 overall post value would be required for a 100% reward for the initial $1 voter, rather than the $16 under 75 / 25. This suggests a maximum capacity of 25% of votes earning 100% of vote value through curation, an increase of x4 from the 6.25% under 75/25.
  • Example 2: Under 50/50 a $16 overall post value would be required for a 100% reward for the initial $1 vote made at 7.5 minutes, rather than the $64 under 75 / 25. (i.e. x4 x4 rather than x4 x16).
  • Example 3: Still zero. But only if the post receives less than $4, rather than $16 under 25 / 75.

All very theoretical!

On to the actual analysis. In this study I will:

  • Analyse the actual curation reward / vote value statistics from the first week of April to derive the distribution of the statistic and use this to consider the "capacity" for 100%+ curation rewards.
  • Repeat the analysis for the first week of May as a check on the April values.
  • Use these distributions to consider how the move to 50 /50 author/curation will change the available 100%+ capacity.
  • Draw conclusions.

2. Curation reward / Vote value distribution from April 1-7

The first analysis looks at the distribution of votes values and curation rewards by the "Curation reward / Vote value" statistic for the period April 1-7.

The data includes all votes and rewards from posts / comments posted in the week April 1-7. An alternative approach would have been to use all votes made in the week April 1-7 but the former approach is (a) simpler and (b) likely to give very similar results, particularly given the one-week reward period - although with the analysis shifted by one week.

Distribution of vote value at each “curation reward / vote value” percentage

DistCapVoteValueApril1-7.png

  • Some votes earn zero curation rewards. Most likely this is an aggregation of very small votes.
  • The peak is at 15% / 16%.
  • 78% of votes earn less than 25% curation with the bulk in the range 12%-24%.
  • 4.2% of votes earn more than 50% curation.
  • 0.8% of votes earn more than 100% curation. I think it's fair to say that the practice of earning 100%+ curation rewards is currently very niche.

Cumulative distribution of vote value at each “curation reward / vote value” percentage

CumDistCapVoteValueApril1-7.png

  • The same results expressed as a cumulative distribution.

Distribution of curation rewards at each “curation reward / vote value” percentage

DistCapCurRewApril1-7.png

  • When looking at the distribution of curation rewards by “curation reward / vote value” percentage the tail is fatter. This is as expected, since curation rewards are higher at higher “curation reward / vote value” percentages.
  • 5.0% of all curation rewards were gained by the 0.8% of votes receiving 100%+ “curation reward / vote value”.

Table of results

TableApril1-7.png


3. Curation reward / Vote value distribution from May 1-7

A repeat of the above analysis for the first week of May as a check that the April values are representative.

Distribution of vote value at each “curation reward / vote value” percentage
DistCapVoteValueMayl1-7.png


Cumulative distribution of vote value at each “curation reward / vote value” percentage
CumDistCapVoteValueMay1-7.png


Distribution of curation rewards at each “curation reward / vote value” percentage
DistCapCurRewMay1-7.png


Table of results
TableMay1-7.png


As can be seen, the May charts are all very similar to those from April.

In terms of our main "capacity" statistic we have:

  • April 1-7: 0.8% of votes receiving 100%+ “curation reward / vote value”, gaining 5.0% of overall curation rewards.
  • May 1-7: 0.8% of votes receiving 100%+ “curation reward / vote value”, gaining 5.2% of overall curation rewards.

4. How will the move to 50/50 author/curation change the available 100%+ capacity?

We cannot say for certain since voting behaviour will change when the economic incentives change. However we can derive a basic "all things being equal" starting point for the discussion.

Under 50/50 author/curation, a vote which currently earns 50% “curation reward / vote value” would earn twice as much, i.e. 100%. So a useful starting point is to look at the level of votes earning 50%+ “curation reward / vote value” from the above distributions. This gives the following comparisons:

April 1-7: 75%/25%

  • 0.8% of votes receiving 100%+ “curation reward / vote value”, gaining 5.0% of overall curation rewards.

April 1-7: "basic derived 50%/50%"

  • 4.0% of votes receiving 100%+ “curation reward / vote value”, gaining 14.5% of overall curation rewards.

May 1-7: 75%/25%

  • 0.8% of votes receiving 100%+ “curation reward / vote value”, gaining 5.2% of overall curation rewards.

May 1-7: "basic derived 50%/50%"

  • 4.2% of votes receiving 100%+ “curation reward / vote value”, gaining 15.1% of overall curation rewards.

In each case the increase in capacity is roughly 5x; not far from the 4x we might expect from the theory in part 1.


5. Conclusions and thoughts

Just my views. Not gospel.

What is the capacity for curation hunting bots to earn more than 100% of vote value (the self-voting level) under 75/25? Will this capacity increase under 50/50?

In the discussions around the EIP there is some concern that a move to 50/50 curation would result in an increase in curation hunting bots which use historic voting pattern data to glean high curation rewards.

The above analysis shows that this practice, or the capacity for this practice, is currently very niche, with only 0.8% of votes earning a “curation reward / vote value” higher than 100%.

Under 50/50 curation, we would expect this capacity to increase. The basic "all things being equal" estimate from the above studies suggests an increase of around 5x, with 4.0% of votes earning a “curation reward / vote value” higher than the 100%.

While this is an increase, even at these levels the practice would remain pretty niche. It's something that may be a draw for minnows and dolphins but there is not sufficient capacity for orcas and whales.

There is also no real potential for the practice to scale significantly above these levels due to the mathematics involved - an increase in SP using this approach will drive down the returns below 100%. It certainly is not going to be the new home for the 40% of passive votes that currently residing in vote bots.


But 0.8% of votes to 4.0% is an increase, nonetheless. So is this practice damaging?

Curation hunting bots which use historic voting pattern data are content agnostic. So it might be assumed that, like vote-bots, an increase in their use would be damaging for the Steem environment.

However curation hunting bots simply amplify the voting that happens elsewhere; pushing votes towards authors that previously received high rewards (Ok, not all curation hunting bots work exactly this way - but the volume will be drawn to the higher reward volumes). So I would assume that if the voting patterns are improved by the EIP, curation hunting bots will amplify these new votes by a marginal amount.


Will increased competition for curation reduce the capacity for the 100%+ returns below the 4.0% of votes derived in the above analysis?

A move to 50/50 would immediately increase the capacity to 4.0% of votes, according to the above analysis, under the assumption of no change in voting practices.

After this we would likely see:

  • Existing curation hunting bots shifting to voting earlier in the reverse auction window, driven by competition between themselves for the higher returns available. This will send more curation rewards back to the reward pool.
  • New SP specifically targeting curation hunting which will also reduce the returns available by increasing competition and a shift to earlier voting. There are only so many opportunities available.

Overall, the capacity is likely to be below the 4.0% of votes quoted. The extent to which the capacity falls will depend on:

  • Where the new voting equilibrium is established;
  • What other opportunities are available to earn (close to) 100% of vote values;
  • The efficiency of the market - not very, given vote bot returns.

What practices might we see to draw new SP capital into curation hunting?

We could see various approaches:

  • Accounts requesting SP delegation for curation bot funds that will return the curation rewards to the delegators (using a second smaller account for liquidity) in return for taking a margin.
  • Accounts advertising curation bot voting trails. Here the accounts gain increased rewards by being the first voter in the trail.
  • Individuals leveraging by taking out SP delegations for fixed costs and using these funds to curation hunt.

Whilst a change to 50/50 would be likely to increased interest in curation hunting, overall the practice is likely to remain pretty niche. In my view it is unlikely to cause the type of damage to the Steem environment currently caused by vote-bots.

None of this post is financial advice! The world may turn out very differently from my thoughts in this post. Do your own research. You'll probably get burned.

Thanks for reading!


Repository:

https://github.com/steemit/steem

This analysis is of data from the Steem blockchain which is an open source project.


Tools and scripts:

gears_blockops_green.jpg

I used the block.ops analysis system to produce this study. Block.ops is an open-source analysis tool designed for heavy-duty analyses of the Steem blockchain data.

You can find the repository for block.ops here:
https://github.com/miniature-tiger/block.ops

The analysis used all the Steem blocks from the period analysed.

The study can be recreated by (once I upload the new analyses to github!):

  • Loading the data for the relevant time period into block.ops.
  • Using the earningsdistribution command from the command line, for example:
    $ node blockOps curationcapacity "2019-04-01" "2019-04-08"

Thanks for reading!

Sort:  
Loading...

Excellent analysis. I find myself wondering how readily one could bidbot their posts sky high for a week, attracting attention from curation bots, then get rid of the bidbots and continue to enjoy the curation bot votes, based solely on past performance.
One might have a dozen curation bots each upvoting their posts because the others will; deliberately engineering a self-fulfilling prophecy.

Thanks Matt!

It would be interesting to see (a) how much in curation bot votes you would get added for say, $100 of bidbot votes, and (b) how long they would continue to upvote.

My guess is (a) only a marginal amount, probably a few dollars, since they have to make x16 or maybe even x64 for high curation returns and (b) only a fairly short period - there must be some regular updating to make sure they're hitting the desired returns.

Getting on voting lists from accounts that vote automatically on a regular set of authors might be a bigger win. Some of those seem to continue forever. Maybe that's something I'll investigate at some point!

This possibility selects for smarter bots. In the scenario you describe at least some of the bots will get a poor return, and possibly all of them.

Since I am a minnow now, I looked through this with great interest... this gives me a little better sense of how curation works, although I am a LONG way from giving anybody a $1 upvote. Thank you for all the work you put in on this!

No problem! I'm glad you found it useful!

Test.

Curation hunting bots which use historic voting pattern data are content agnostic

I would say not entirely, although I am mostly agreeing with you that they amplify or perhaps just accelerate existing patterns. To the extent that the votes they are looking to predict are content agnostic, they are too, but to the extent the votes they are looking to predict are not content agnostic then the curation hunting bots aren't either.

Example:

  1. Author X consistently gets high rewards due to valuable content
  2. Author X consistently gets high rewards due to close-to-break-even purchased votes.

These are very different IMO and likewise whether curation bots are hunting #1 or #2 is also very different.

#1, I would view as doing a service two ways:

  1. Curation hunting bots and competition between them returns more curation rewards to the pool
  2. Since it surfaces the content faster, people can then use rewards as a real time and trusted content discovery method (trustworthy because there are real resources betting on it).

If free downvotes work, there may not be much if any amplification (rather vote acceleration, as I noted above), because if the content ends up overrewarded in toto (curation bots plus later votes), it may attract some downvotes. We'll have to see if that actually happens though.

Hi @miniature-tiger!

Your post was upvoted by @steem-ua, new Steem dApp, using UserAuthority for algorithmic post curation!
Your post is eligible for our upvote, thanks to our collaboration with @utopian-io!
Feel free to join our @steem-ua Discord server

Hey, @miniature-tiger!

Thanks for contributing on Utopian.
Congratulations! Your contribution was Staff Picked to receive a maximum vote for the analysis category on Utopian for being of significant value to the project and the open source community.

We’re already looking forward to your next contribution!

Get higher incentives and support Utopian.io!
Simply set @utopian.pay as a 5% (or higher) payout beneficiary on your contribution post (via SteemPlus or Steeditor).

Want to chat? Join us on Discord https://discord.gg/h52nFrV.

Vote for Utopian Witness!

It's not a theory, curation hunters exist.

You are spot on... let that HF drop!!!

Congratulations @miniature-tiger! You have completed the following achievement on the Steem blockchain and have been rewarded with new badge(s) :

You distributed more than 9000 upvotes. Your next target is to reach 10000 upvotes.

You can view your badges on your Steem Board and compare to others on the Steem Ranking
If you no longer want to receive notifications, reply to this comment with the word STOP

Vote for @Steemitboard as a witness to get one more award and increased upvotes!