Math of Steem: The HF20 vote floor and curation trails
One of the lightly-advertised changes in HF20 is that the voting calculation will change a little bit to adjust for the fact that SP-based bandwidth is going away and many new accounts will start with 0 SP. This will lead to a lot of votes lower than the dust vote threshold, and the system needs to handle that somehow; what the developers have decided to do is just not count the first fifty million rshares of everyone's vote.
50,000,000 looks like a big number, but in the case of rshares it really isn't. Right now that's a vote equal to about $0.000067. So if you're voting on your own in reasonable percentages, this really won't make any difference to your Steem experience, and I'm sure that's what the developers were thinking about when they did things that way.
However, it is a tiny penalty placed on those who vote more often, and one place where that can really matter is in curation trails, where people are often giving extremely small votes already with the expectation that they will count. I wanted to take a look at what's generally the most-notable curation trail on Steem, the one belonging to @curie.
If you've gotten a Curie vote, you know very well that it's not just one vote. The curation trails votes just pour in, and if you've got Ginabot notifications it can be pretty overwhelming. A main-line Curie vote brings over 800 votes with it, and the subsidiaries are still 125 or so. Even their very tiny votes have 50 followers.
(I made a units error in the original version of this post; it has been corrected below.)
The effect starts to be a little bit noticeable at that scale. Every main-line Curie vote would lose about five cents total under HF20 with current prices. Just looking at yesterday, Curie handed out 22 votes greater than 10%, 52 votes at single-digit percentages, and 144 votes at 1% or less. @curie by itself will hand out almost 11 billion fewer rshares every day under the new plan - approximately a single vote from an account with 270 SP, a little less than two cents.
But when we include the curation trails the number balloons: that's 31300 votes being used to make up a full day's Curie payout, which is a startling 1.565 trillion rshares lost to the floor. That's equivalent to a 100% vote from a large dolphin with about 38,650 SP, or around $2.17 less every day that Curie collectively will be handing out in rewards.
Maybe that's still not a huge deal, but it's no longer completely trivial. A little bit of that will come back in increased rshare value, but most of it will be distributed to accounts that vote with higher percentages less often and thus lose fewer rshares to the floor. Since that's most of us maybe you can see that as a good thing. But in general curation trails are going to lose a small chunk of their value in HF20, and I'm not sure anyone really is paying attention.
Curious.
Didnt know that curie's trail is so crowded.
Nice post.
I am curious about the exact details. Is there already a formula sheet available like this -> https://www.steem.center/index.php?title=Rewards:Formulas for the HF20?
Not that I know of. For this issue it's pretty simple: instead of rshares being Vs(pu/100%), effective_rshares will be Vs(pu/100%) - 50,000,000 (minimum 0).
okok got it.
The github contains some of the formulas but it is pretty scattered.
For the current power used,
where w_p is the weighted power, the 0.49%/50 in some sense corresponds to the price you need to pay to cast a vote (in terms of your VP). I am curious if this formula will still remain or if the 50,000,000 is meant to replace it.
They're also changing a lot more about how the voting system works in switching to voting mana, so I have no idea. I'm guessing that .49/50 is there to handle some weirdness that comes at very small vote sizes, and/or stop people from exploiting vote quantization, which is something else I don't understand very well. I don't think it's related to dust votes.
The 0.0049/50 is VP that is not represented in the value of the vote since the value of the vote is given by w_p times a constant. So if you are voting a lot ot low weigthed votes instead of high weigthed votes then your VP is going to be lower.
In addition, if you would want to optimise the votes that you can get out of an account in a single date then this 0.0049/50 ensures that there is a max amount votes you can cast. If you cast more votes then that then the total value of votes cast will be lower. This has something to do with the weirdness that comes at very small vote sizes. :)
I hadn't even thought about the effect it would have on curation trails. That is significant if it's going to take away the equivalent of a whale vote from just the effect on Curie.
Ugh, I screwed up units somewhere here. It's not really that high, trying to work the math again.
I think it's fixed now. I'm really not clear how I ended up off by multiple factors in multiple spots including in one case by a factor of 20, somehow. There aren't any 20s. But I went back and reviewed all of the units from scratch and I'm more confident in them now.
I guess I made a factor of 1000 mistake in one direction and a factor of 50 mistake in the other, at the same time. I shouldn't make these posts before breakfast.
Ha ha, that's alright, man. Thanks for rechecking the math. I guess we're not in as much trouble as we thought. :)
Something to think about. I follow a curation trail and receive a lovely of votes from one. I’ll have to see whether I want to change tactics and manually look members up a couple times a week.
Posted using Partiko iOS
What I would find more interesting than this is what kinds of accounts are part of trails but are going to lose their votes. Are they active accounts, are they 'set and forget', are they curation miners, are they active in other ways or, are all of their votes automated and they aren't actually present on platform at all.
I wonder how many inactive accounts are still voting on trails. Hard to put a number on I guess considering they are still casting votes but perhaps, if they haven't posted, commented or cast any other non-trail vote in some period of time it could give a ballpark number.
@abh12345?
SteemAuto appears to avoid casting votes that will be dusted. That's why the trail numbers are different for different sizes of Curie vote. So nobody should lose their entire vote from this effect, but they might lose a significant percentage of it. How much may depend more on how @mahdiyari decides to handle it than on the HF20 changes.
That's an interesting question, though.
I know that I get tiny votes worth nothing from accounts that haven't done anything else for 6-12 months. If the attrition rate is as high as they say, I would suggest that there are many small accounts that have tried to maximize with autovoters but are now dead.
Taken from: https://steemit.com/utopian-io/@crokkon/how-will-the-hf20-dust-vote-changes-affect-the-vote-values
I would expect that may of these accounts on @curie's trail (and the ones front-running bots) don't have 120SP and have around the 15 SP 'starter' amount. These accounts will need something like a 15/120 = 8% vote percentage (if at 100% VP) for the rshares to register.
It's a bit of a pain to work out who will need to alter their approach, but as long as you have 120 SP + and register a 1% vote (at 100% VP), it will count, with less rShares lost % wise, the larger the vote being cast.
Ah yes, this is something we are actually currently trying to brainstorm at Curie.. how we can make changes to ensure those votes aren’t lost, seeing that most are following with a low percentage. Any suggestions are greatly appreciated 😉.
I have a suggestion. (not my area)
Is it possible that the trail will only include voters who pass the threshhold? Or, a voter sets their vote to only trail every second or third curie vote so they can increase the % of the vote without draining SP. I don't know if it is possible at the moment but it seems that it should be simple enough to add for the auto voter platform. What it could mean is that even though there are less accounts in the trail each vote, the vote amount itself won't be adversely affected that much.
1.2 SP at 100% VP and with 100% Vote Weight will count - they will just take a hit with regards to rShares earned.
The accounts receiving the votes don't need to worry, the numbers are tiny. It's the accounts making the votes who may not earn anything, if this is their approach. (And if this is their approach, meh).
I don't think @curie need worry, not much will be lost to the 'voted', it will be the tiny accounts on auto-pilot that will lose out the most.
I actually think this is less of a problem than people think but this is my opinion. The starter amount voters aren't necessarily very active accounts and even though their little votes add up over time, they themselves as accounts may cause more harm than good overall. Also, I stil wonder how many dead accounts are parts of trails. this means that they will vote forever and even if they earn curation it will potentially be locked up forever too.
Same. It's a minor issue to very low SP accounts, most of the 'engaged community' have above the threshold, even with a 1% vote going down.
Maybe they plan to return in 10 years time :) The account may not have grown though due to HF 20 - shame :P
Yep. I feel sorry for people who expect to get something for nothing.
I think most people are clueless about HF20, including myself. I do take time to read explanations that people write about it like this one, though I'm not really sure if something needs to be done or not except that I need to upvote posts 15 minutes after when it's implemented.
That's the main change that's really important from a user's perspective. For this particular change it's good to re-evaluate any curation trails you're a part of, especially if you're giving them very small votes. You'll get more value by using voting power in larger chunks.
You got a 9.46% upvote from @upmewhale courtesy of @bidbot.credits!
Earn 100% earning payout by delegating SP to @upmewhale. Visit http://www.upmewhale.com for details!
This post has received a 13.87 % upvote from @booster thanks to: @bidbot.credits.
To listen to the audio version of this article click on the play image.
Brought to you by @tts. If you find it useful please consider upvoting this reply.
You got a 18.60% upvote from @brupvoter courtesy of @bidbot.credits!