STEEMIT: Facts about Curation Rewards

 A beginners guide to making money discovering quality content before everyone else.

Voting on Posts

Users that have made a multi-year commitment to Steem by converting their STEEM to Steem Power have the ability to vote on posts. The weight of each vote is proportional to how much Steem Power you have divided by the average number of votes you cast in the last 24 hours. If you have cast less than 20 votes in the last 24 hours then 20 will be considered your average.

Rewarding Curation

Steem wants to reward those who sift through the mountains of uncurated posts to find the diamonds before anyone else. Steem has designed its curation reward system to maximize rewards for voting on popular content before it becomes popular. This article will dive into the mechanics of how this is accomplished so that those hoping to game the curation rules to maximize their own rewards can do so.

How it Works

Every time a post gets paid, 50% of the payout is directed toward those who contributed the most to increasing the posts payout. For those who don't care how it works and just want some basic rules to follow, here are some guidelines:

1. Only vote on posts that you believe others will also vote for.
2. Vote as early as possible after the content is posted.
3. Don't vote on content that is already popular
4. Acquire as much Steem Power as possible
5. Vote for less than 50 posts per day
6. Avoid downvoting

The less Steem Power you have the more important it is for you to focus your votes on a
few good posts rather than spreading your vote thin. In particular, users with small amounts
of Steem Power should:

1. Vote on content others haven't voted for, to maximize the percentage increase in voting
2. Vote on fewer posts, the more you vote the more diluted your Steem Power.

Challenges with Rewarding Voting

The saying "you get what you pay for" has never been more true. In this case Steem must be
"careful what it pays for" or we might not get the result that we want. If Steem simply paid
people for voting then tech-savy individuals could simply vote for everything to maximize
their payout. This type of behavior would add no new information and therefore no value to
the platform. The reward algorithm must be designed to reward information while preventing
gaming and automation.

Reward Percentage Increase in Voting

The algorithm has been carefully designed to minimize the impact of automated voting
algorithms. In particular, we want to reward voting on posts that the network
has the "least information" about. This means that being the first person to vote for
a post gives you the greatest possible curation weight for your Steem Power.A human has a better chance of predicting the success of a random post with no votes than
a computer algorithm. At best a computer algorithm could use historic performance of the
poster to guess. If an algorithm is used, it will have to be a sophisticated algorithm
that actually does add some value. After all, the algorithm isn't able to vote on everything
and gets exponentially weaker more it dilutes its vote.Someone who adds 1 vote to a post that already has 99 votes will get a weight of
.01. On the other hand, someone who adds 99 votes to a post that
already has 99 votes will end up with a weight of 25.The order of voting matters. Given 4 users with 1 vote, and 1 user with 4 votes the
payout received by individual voters can be dramatically different. If the 1 vote
users vote first, then their weights will be:

 Vote:      1,   1,  1,  1,   4

  Weights: 100,  25,  9,  6,  25

  Percent: 60%, 15%, 5%, 3%, 15%

If the order were reversed then the weights would be:

 Vote:      4,  1,  1,  1,  1

  Weights: 100,  4,  3,  2,  2

  Percent: 90%, 3%, 2%, 2%, 2%

Reward Vote Concentration

Human voters can only process so much content per day. A script that votes on everything is
guaranteed to have a much smaller percentage of the winning posts than someone who applies
some human intuition. This vote concentration is rewarded automatically by paying out
rewards based upon percentage increase in total votes (not total payout), with one small caveat: the first voters
can have very large percentage increases with very small stake.Imagine someone who has 4 accounts with STEEM Power of 1, 2, 3 and 6. If they vote consecutively,
then each vote after the first would see a 50% increase, with the first vote seeing 100%. If we
were to sum the weights based solely on (percent increase)2, the attacker would start out
with 1.75 using this strategy compared to starting out with 1.0 by voting from one account with 12 STEEM Power.Ideally, it should always be better to vote with 1 account than to divide up your stake and vote
from multiple accounts. To achieve this we multiply the (percent increase)2 by the Steem Power of
the vote. This gives the following vote weights:

Multiple Accounts: 1 + 1 + 1.5 + 3 => 6.5

 Single Account: 12

The conclusion is that someone voting with all of their stake at once gets twice the final payout of the same
individual voting multiple times through multiple accounts. It is true that the rich get richer when
viewed from the perspective of an individual post, but when viewed as a whole the playing field is
much more even. Regardless of how rich someone is, they can only evaluate so much content. The rich
individual also has the most to lose from poor voting.

The Math

The network will automatically divide the 50% payout among all users who voted for a post on a
pro-rata basis using the following equation to measure the weight of each voter:

let total_vote_reward         = 50% of the content reward

let steem_power               = the total Steem Power of the vote

let current_total_steem_power = the combined Steem power of all past votes

let new_total_steem_power     = current_total_steem_power + steem_power;

let vote_payout_weight        = steem_power * (steem_power/new_total_steem_power)^2

let total_vote_payout_weight  = sum vote_payout_weight for all votes on the post



let vote_payout = total_vote_reward * vote_payout_weight / total_vote_payout_weight