Whaleshares Logo

Curation Maximisation Explained - Part 2

scipioPosted for Everyone to comment on, 5 years ago9 min read

Curation Maximisation Explained - Part 2

math.jpg

Hello, Whaleshares!

In yesterday’s post ( Curation Maximisation Explained - Part 1 ) we looked in=depth at the Curation Rewards formula, and discovered - via 2 examples - that it’s better (seen from the perspective of increasing your own curation rewards at least) to vote on 0 WLS posts than it is to vote late on high payout posts (meaning, after a lot of other rewards came in and the post value is not expected to grow much further after you voted).

We also saw that the maximum of combined curation rewards on any given post is 25% of its total payout value.
Meaning, that if a post has, for example, a PT (final Payout Total) value of 1,000 WLS, a maximum of 25% = 250 WLS will go to all combined curators that voted on that post. If however early voting happened, so before t=1 at/after 30 minutes, either by the author itself or others, than parts of those votes are removed form the curation rewards portion and into the author rewards portion.

However, that “25%” on itself has nothing to do with your vote. We did say, that if you vote on a 0 WLS post and nobody else votes after you, that indeed you get 25% of your vote value as curation rewards, because if your vote is the only vote then that also means your vote is equal to all combined votes. In case you vote smart, however, and many other (big) votes come in after you voted, then you can have a relatively big “weight” among all those votes leading to a situation in which your “share” of the combined curation rewards could be bigger then your own total vote!

Let’s have a look at another example to see how that could work…. And as a brief reminder, again here is the curation reward formula (written in 2 forms, and explained again per variable):

-1-

cur = K * v * t * sqrt(s)

where
K = [1 - (sqrt(1-k))] / 4k
and
k = v / P1 = v / (P0 + v)

and
s = PT/P1

and therefore

-2-

cur = 0.25 * sqrt(PT) * (sqrt(P1) - sqrt(P0)) * t

cur is of course your curation rewards.
PT is the total, FINAL value of a post, right before the post payout. The Payout Total
P1 is the value of a post immediately AFTER YOU have added your own vote to it
P0 is the value of a post RIGHT BEFORE YOU have voted on it
v is the value of your own vote, in WLS
t is the “time” value, which is 0 the second a post is published, then grows linearly between 0 and 1,800 seconds (= 30 minutes) to 1, and stays 1 until post payout.
sqrt is short for square root
k has to do with the scale of how much your own vote has grown the post value at the moment you have voted on a post
K is pretty difficult to understand as it’s not directly in the original curation formula, but it’s a hyperbole always ranging between 0.125 and 0.25. The higher, the better, seen from your own perspective.
s has to do with the scale growth of the post payout after you have voted until post payout.

Example 3: Turbo Boost your vote value

Let’s say at exactly 30 minutes (t=1) after it was published you stumble on a very interesting post from a new author that hasn’t got any stake but seems like an experienced blogger, and you think the post is really good and will grow to (again, example-wise) 1,000 WLS as PT, but it’s right now at 0 WLS (P0). Your own vote is worth 10 WLS (v=10), so P1=10 as well. Afterwards, because you let a few other people know about the post as well the post value grows by another 990 WLS to 1,000 WLS in total at PT.

Now have a look at what this means:

cur = 0.25 * sqrt(PT) * (sqrt(P1) - sqrt(P0)) * t

ergo:

cur = 0.25 * sqrt(1000) * ( sqrt(10) - sqrt(0) ) * 1
cur = 25.0

So: in this specific situation, you voted with “only” 10 WLS but you got 25 WLS in return as curation rewards. How is that possible, as that is 2.5x the size of your own vote value?

To begin with, the post value grew 100x after you voted 10 WLS on it: s = 1,000 / 10 = 100. In the formulasqrt(100) leads to a factor of 10 influencing your curation rewards. Also, because you were the first to vote on it the k-value is 1: k = v / P1 = 10/10 = 1, meaning K=0.25, being the maximum value. (If you’re the first to vote, K=0.25, the maximum.) And as cur = K * v * t * sqrt(s), in this case: cur = 0.25 * 10 * 1 * sqrt(100) = 0.25 * 100 = 25

Another way looking at this: curation_performance = cur / v and in this case: curation_performance = 25 / 10 = 2.5, meaning you indeed get 2.5x as much curation rewards as your own vote is worth. That’s pretty good!

Example 4: Less than 100% voting for Big Accounts vs Small Accounts

In example 3 you found a 0 WLS post that you were the first to vote on with 10 WLS and in the end the PT grew to 1,000 WLS, where the scale growth s=100, leading to a staggering 2.5x Curation Performance. Of course, that situation is very rare, with K=0.25 and s=10=sqrt(100) at t=1 (after 30 minutes). However, there are many posts that have parameters being close to example 3.

@haejin for example posts about 10x per day, and due to his large amount of Whalestake, let’s say that on average his posts have a PT of 250 WLS. For one reason or another, @haejin does self-vote but only after a few days, ergo at t=1 instead of t=0. That gives others an opportunity to “front-run” his self-votes and claim some curation rewards from it (also by giving @haejin more Author Rewards — there’s of course nothing wrong with voting on anybody else’s content including that of @haejin).

haejin.png

I took a snapshot of a post that @haejin published about 2 hours ago, now 2 people have voted on it and the current value (P0) is at 3.19 WLS. Let’s say I expect PT to grow to 250 WLS after a few others (but excluding myself!) and @haejin himself have voted on it. My own full vote is about 42 WLS right now. Let’s say I cast a full 100% vote on his post, leading to an expected PT of 250 + 42 = 292 WLS. What would that mean, for me, curation-wise?

t = 1
P0 = 3.19
If I vote right now with v = 42 then P1 = 45.19, and as said,
the expected PT is then 292.

cur = 0.25 * sqrt(PT) * (sqrt(P1) - sqrt(P0)) * t and substituting values to cur = 0.25 * sqrt(292) * (sqrt(45.19) - sqrt(3.19)) * 1 so cur = 21.088. That’s still about half of my own vote value (cur_perf = 21.088 / 42 = 50.21%), which is twice as good as voting on posts at 0 WLS on which I’ll be the only one voting (!!!).

However, let’s say I would vote only a fraction of a full vote, for example only voting at 0.56 WLS. How about if I’d cast a 0.56 WLS vote on this exact same post by @haejin, instead of a full vote worth 42 WLS? What would that mean in terms of curation performance? In that case, the expected PT = 250.56 but P0 is still at 3.19 and P1 would there be 3.75 WLS.

Ergo: substituting values in cur = 0.25 * sqrt(PT) * (sqrt(P1) - sqrt(P0)) * t as cur = 0.25 * sqrt(250.56) * (sqrt(3.75) - sqrt(3.19)) * 1 so cur = 0.595, meaning the curation performance is 0.595/0.56 = 106.2%.
Interestingly, even though P0 > v (the value before I voted is bigger than my own vote) meaning K is sub-optimal, because now (due my smaller vote and especially the relative low P0 value as compared to the expected PT) the expected scale growth s is large:

Scenario -a-, voting with v=42 at P1=45.19 and PT=292: s = 292 / 45.19 = 6.46 and sqrt(s) = 2.54
Scenario -b-, voting with v=0.56 at P1=3.75 and PT=250.56: s = 250.56 / 3.75 = 66.82 and sqrt(s) = 8.17

Now, remember (as a rule of thumb): the K value in cur = K * v * t * sqrt(s) is at minimum (if you vote last and an enormous amount of large votes came in before you) ~= 0.125 and 0.250 max, the ratio 0.25/0.125 = 2. But even though in scenario -b- the K-value was less good than in scenario -a-, now the ratio factor sqrt(s) 8.17 / 2.54 = 3.21.
And as 0.50 (worst-case K-value “loss”) * 3.21 (scale-growth ratio “win”) > 1, the curation performance of this “small vote” still outperforms the full-vote.

What can we conclude from all this?

That maximising your curation rewards is pretty tough, as of course NOBODY is going to calculate their potential curation rewards manually before voting every time. Neither do I, I only explained how curation rewards work by doing some math. ;-)
It is possible to use a voting bot that does a lot of computations and casts votes according to your own rules in order to maximise your own curation rewards. I might deploy one for myself, but surely NOT as a public community service or something — I don’t like voting trails for example, never used one myself. Others might, I won’t.

Again feel free to ask questions, in the comments!

For now, thanks for your time!

Kind regards, @scipio

Sign Up to join this conversation, or to start a topic of your own.
Your opinion is celebrated and welcomed, not banned or censored!