The Tale of the 5 Brothers -- A Voting Power Parable
People often notice the many similarities between me and Jesus. Today, I will attempt to explain the voting power changes coming on the 13th in a parable, just like Jesus would.
Going into this, I want to correct an error in my earlier post on the subject. In my earlier post, I misunderstood some graphs on the white paper and assumed that the vote target was a "sweet spot" and that going above or below it would result in a penalty to voting power. This, in fact, is not true. You are only penalized for casting too few votes, not for casting too many. If you cast more than the target number of votes, your decreased voting power will be entirely offset by your increased number of actual votes.
You can read some of the discussion that led to this post here and here
The 5 Curating Brothers
Imagine for a moment that there are 5 brothers. These five brothers do all of the curating on steemit. They all have the same amount of steem power. In this theoretical steemit system, the total daily reward pool is $8,500.
Andy devotes the most time to his curating duties. He casts the target 40 votes per day.
Ben is not as good as Andy at managing his time and his votes, but he still tries. He casts 30 votes a day.
Carl doesn't have very much time to delegate to curating. He casts 20 votes a day
David Has even less time. He casts 10 votes a day.
Frank barely has any time to devote to curation, he casts 5 votes a day.
In total, 85 votes are cast every day to award $8500. Each vote, therefore, is worth $100.
Brothers Vote Value At the 40 Vote per Day Target
| X | votes cast per day | Median Vote strength | Adjusted Votes Per Day | total value cast per day |
|---|---|---|---|---|
| Andy | 40 | 1 | 40 | $4000 |
| Ben | 30 | 1 | 30 | $3000 |
| Carl | 20 | 1 | 20 | $2000 |
| David | 10 | 1 | 10 | $1000 |
| Frank | 5 | 1 | 5 | $500 |
| Total | 85 | 1 | 85 | $8500 |
Note that, for a fixed amount of steem power and reward pool, Andy is unique in that he cannot earn more money by voting more. He wouldn't lose money, but the change to his voting power would make any vote above 40 a wipe. He actually can lose money by clumping his votes. Ideally, at 40 votes, he should be able to cast every single one of his 40 votes at 100% power by spacing his votes at 36-minute intervals through the day. If there was a super-Andy who cast 80 votes per day, his adjusted vote value would be the same as Andy's, because he would be unable to cast his full 80 votes and also wait long enough to cast them all at 100%.
Brothers Vote Value at the 5 vote per day target
| X | votes cast per day | Median Vote Value | Adjusted | total value cast per day |
|---|---|---|---|---|
| Andy | 40 | .125 | 5 | $1700 |
| Ben | 30 | .166 | 5 | $1700 |
| Carl | 20 | .25 | 5 | $1700 |
| David | 10 | .50 | 5 | $1700 |
| Frank | 5 | 1 | 5 | $1700 |
| Total | 85 | .29 | 25 | $8500 |
Notice that, because they are above the vote target, Andy, Ben, Carl, and David all have median vote values below 1.0 Just like Super Andy in the previous example, the "extra" votes that they cast above the target threshold attenuate their voting power. Andy could bring his voting down to 5/day (just like frank) and be able to cast 100% votes, but it wouldn't really make a difference. He is not being penalized for casting too many votes.
Note here that the monetary value of a single 100% vote did go up.
To begin with, it was
8500 / 85 =100
After the target change, it is
8500 /25 = $340
In this particular case, it went up just over 200%. However, this increase was insufficient to make up for the 87% decrease in the number of votes that could be cast. The farther away each brother was from actually casting the maximum number of votes, the less his average influence was effected.
To use shennanigator's model, if you could cast 40 100% votes per day worth 10 cents each before the change, after the change you would be able to cast 5 100% votes worth 34 cents after the change.
Here are the before and after numbers for the total $ influence exerted by the brothers.
| x | 40 vote target | 5 vote target | % gain/loss |
|---|---|---|---|
| Andy | $4000 | $1700 | -57.5% |
| Ben | $3000 | $1700 | -10% |
| Carl | $2000 | $1700 | -15% |
| David | $1000 | $1700 | +70% |
| Frank | $500 | $1700 | +240% |
Lowering the Bar Does not Improve Participation(it just improves how many get over the bar)
Biophil wrote in my previous thread
It just depends on total voter participation; if participation goes up (a good thing), the increase won't be 8x. If it goes down (a bad thing), the increase will actually be more than 8x. It's very difficult to predict.
This is somewhat misleading. Imagine a classroom with 40 students in it. At the end of the year, 4 students are scheduled to be awarded a certificate for perfect attendance. On the very last day, the principal comes into the classroom and announces that, in addition to the 4 students who actually have perfect attendance, 10 students who have been absent 2 days or less will also get the certificate.
By doing this, the principal has not, in fact, increased attendance. He has merely increased the superficial appearance of attendance by rewarding imperfect attendance the same way he rewards perfect attendance. In the same sense, the 5-day target does not increase participation. It merely gives the same reward to lower participation as it does to higher.
In fact, it disincentivizes high participation. In the new system, there is less reason for andy to take the time to read and vote on 40 posts. He can still do so, but why should he? He can be like frank, put in much less work, and get the same level of influence.
I heard the new voting target is helping fight bots... how is it doing that?
It's really not. At least, not directly. It does hurt bots (because most bots are Andys, though I would argue that most Andys are not bots). Basically, if you're a high power frank, this new target helps you at the expense of all andy's, bots included. When we say this decreases the influence of bots, what we really mean is that it decreases the influence of all very active curators.
Which brother are you?
If you look at the charts above, it becomes clear that there are two factors that influence how the change to a 5 vote target will affect each individual voter.
- How close are you to optimal voting right now? If you're an andy or a ben, you're probably going to see a decrease in influence.
- How close is everybody else? If you're currently a more optimal voter than average, this will definitely hurt you. If you're a less optimal voter, this will definitely help you.
In a world where every poster was an andy, this change would have precisely the type of cosmetic effect @shenanigator and others have described. If everyone on steemit were a Frank or a Dave, and there were only a few Andy's, this vote would decrease the voting power of the andy's significantly.
At the end of the day, what you need to take away from this and my other post on the topic is that this change is not cosmetic, or a "bug fix", or a way to give users two dollar bills for 8 quarters. It is a very real re-distribution of influence with very real consequences to active curators. We should look long and hard at the effects of this redistribution when it happens, because its very possible that some of the best curators here will, to coin a phrase, get Franked in the Andy.
Good analysis but your example seems to be revised with my data HERE. According to the data, a high power Frank seems unrealistic, and we have a high power Andy. You can check many on the list are big stakeholders and they have high IPR (measurement of utilization of power) with many voting casts.
A high power Andy still can have broad but flatter influences by casting below-100% votes.
Ned is a high power frank. Also Dan's account @dan. Also @jamesc and @val-a . NOt talking trash on them, because obviously they have other things to do besides curate content (like run the company), but according to the amount of votes they cast, they're definitely franks. If they continue to vote according to their current habits, they'll have significantly more influence after the change.
By high power frank, i mean someone with a lot of steem power, currently far below the 40 vote daily threshold.
Just as a side note, I discovered when i was looking for high powered freds on the richlist, that there are a very surprising number of high SP accounts that are "dead freds" ... they simply don't curate at all. There are 2 in the top 10 -- the biggest is @ben who has more SP than bernie sanders. and like 4 more in the top 20.
This is surprising to me... it seems like passing up on free money.
Yes there are many high power Frank too. However, ten (out of over 80,000) powerful Andys already have 60% of total value cast per day, and their multiplying factor between total value cast and actual SP share is 2.67 on average (while top 10~20 has 1.43 and top 40~50 has only 0.75). I guess the cause is a high number of voting cast with the combination of superlinearity. I am really curious the results of this change and will post about it after the next hardfork.
After reading your post and this comment, btw, I think I have one primary criticism. I suspect @smooth had effectively the same problem with the post, which contributed to much of his expressed rancor.
You assume, without backing it up with evidence, that a difference in results automatically speaks to a biased or unfair game. It is possible that you are correct, but without evidence, it could speak to one of two things:
Now, in a very technical sense, these are, in fact, the same thing. The fact that the game being played is biased toward a particular strategy is what makes the strategy objectively good. But when you talk about bias and unfair advantage, you imply a particular causation. Specifically, you imply that the rules of the game were designed to give specific players an advantage. However, the very opposite might be true.
For example, imagine you are watching a game of no-limit hold'em. One professional player is playing against 8 individuals who are recreational gamblers. Over even the medium term, absent some very unfortunate luck, the professional player will almost always be winning.
Now, you might say "This game is rigged! The professional player is winning all the money". And, in a sense, you would be correct. The game of NL Holdem is, in fact, biased to reward the professional's strategy (he tightens up out of position, cbets, punishes limp-callers, etc) and penalize the recreational gamblers.
But when you say "the game is rigged for the pro" you imply that the pro's strategy predates the game. This is untrue. The game came first, and the pro developed a winning strategy. It would be pretty unfair to change the rules in the middle of the game to attempt to cancel out the Pro's edge. The rules were there to begin with, and he put the cognitive effort in to discover and refine an effective strategy.
Also, at the end of the day, its pointless to attempt to revise the rules to deprive the pro of his advantage. Because the whole reason he has that advantage in the first place is that he's better at everyone else there (or willing to work harder) at developing an effective strategy.
@smooth and the other whales who have what you would consider a disproportionate influence have it because they invested a lot of time, effort and thought into the development of a curation and posting strategy that gives them that disproportionate level of influence. Isn't that the whole point of making curation a game?
I agree that they are active curators (with large amounts of VESTS) and their influences are due to their efforts. So one important index in my stats is IPR, which measures utilization of Steem Power. However, we need to look at the Max and Stdev of their votes simultaneously. While @blocktrades has very low Max and Stdev, which means his votes are dispersed to various writers, @smooth 's votes are concentrated to few people. That is, blocktrades has disproportionate power for many while smooth has disproportionate power for few, IMHO.
In addition, this analogy overlooked the basic things--super-linearity (votes on post is squared to calculate rewards).
Yes, it did, but im not completely up to speed on that. This model assumes all the brothers vote on brand new posts, and none of them vote on the same posts.
I actually have really strong opinions on the superlinearity issue, but the main scope of this post was to show that, contrary to what many many people on the platform believe, this is not merely a cosmetic change. Users are not giving up 8 quarters and getting two dollar bills. This is a genuine redistribution of influence.
You can't not have superlinearity in any system that successfully combats Sybil.
Good post! Yeah, it all seems right enough. I'd love to know what the actual proportions are between our Andys and Franks.
One thing you left out of your model is the fact that Andy can cast most of his votes at 12.5% slider power, and then now and then cast a mega-vote at 100% power if he wants. This wouldn't change any weekly averages, but he could grab a bunch of influence in spurts. Still, he was almost certainly a lot better off before the change.
Yeah, I'm seeing that now. The devs are measuring voting-quality by the absolute amount of voting power that gets injected into the system, but that's probably the wrong metric.
One more thing that your model leaves out is that the actual posts that people vote on may change, so we could see massive redistributions of posting rewards. Maybe the high-paying blockbuster posts will get voted higher, maybe they'll get voted lower. Very hard to predict, imo.
Well, you have to keep in mind that its not sheer numbers of franks or andys that matter, its the SP weighted porportion of franks and andys. Take a look at the top 50 of the richlist on steemstats and you'll see that many of them are infrequent curators with near 99% power pretty much all the time.
The "ned" effect as I called it in my other post, is going to be pretty significant, because there are a lot of whales like ned who are too busy to be Andys or even Bens.
Now a big question is are they going to "dial down" their 100% votes. Given that theyd be acting against their own interest, its hard ot believe they would.
if it is the latter, we are all fked.
Franked in the Andy is just good writing!
I know, right... I messed with the whole naming scheme to accommodate that line (the last guy was originally Ethan)
I can't imagine a better reason for editing a post.
I think the result may well be that more active curators do lose influence as you suggest. I don't really see how or why this is a good thing. We will have to see what happens but I wouldn't be surprised if this needs to be reversed like the 12 hour payouts.
Well, you have to keep in mind that there must be some threashold where we stop rewarding increased participation, otherwise bots will dominate the curation game. Even now, super andy's are "penalized" in the way you describe. And even if we allowed superandys to vote 80 times, there would be super-duper andys casting 160 votes that we would be "penalizing".
I use the word "penalizing" in quotes becasue its not really a penalty. For example, to borrow from the classroom analogy I cited above, when the principal decides to give perfect attendance certificates to children who have missed one day of class, you wouldn't really say the children who actually had perfect attendence had been penalized, except inasmuch as their award was now devalued.
I understand what you are saying but I just think that the previous voting threshold was actually better. Im not advocating having no threshold. I just think this is a change in the wrong direction. Yes people can change their voting weight but it still changes things. They are essentially letting people use more of their "rewards" on a single post if they want to. The problem is that is not how it will be perceived. When implementing things like this the developers must always take account of human psychology.
This made me laugh so hard. Solid post, I think it's a really good analysis. I agree that the best curators could be hurt by this and that it should be scrutinized.
One thing I think you should have mentioned as a benefit - now users will have more flexibility in deciding the power of their vote.
Edit: After reading through some comments I see biophil already mentioned this.
haha.. thanks.... im glad out little debate turned into something productive and informative to a lot of people.
Yeah, that worked out well. And my vote for your post may have been my largest curation reward yet, so that's nice.
Well explained and illustrated!
Is there any tool to see the number of votes I cast per day on average? When the slider came in, I started experimenting with it dialing voting power up and down, but I've noticed my voting power is staying close to 100 % even with full votes, so at the moment I am nowhere near close to being an Andy. Knowing how much I am voting should indicate how much I will need to dial back my votes after the hard fork to keep an optimum voting power (around 80% I think I recall reading somewhere).
I know steemstats.com will tell you the number you cast in the past 24 hours.
Also keep in mind that there is no "optimal" voting power. You aren't penalized for voting too much, only too little.
If you liked this post, check out some others by YT,
https://steemit.com/steem/@sigmajin/what-every-user-should-know-about-the-upcoming-change-in-voting
https://steemit.com/abuse/@sigmajin/lol-flaggaments-read-of-you-re-the-victim-of-abusive-flagging
https://steemit.com/censorship/@sigmajin/censorship-and-sensibility
that was interesting :) i followed you ..perhaps u look at my blog and u find some graffiti that u like! <3 kalipo
"People often notice the many similarities between me and Jesus..."
perhaps the best opening line of any post i've read here yet. :-)
It looks like the vote rebalancing has been taken out of the new steem version.
https://steemit.com/steem/@steemitblog/steem-0-14-1-released-hardfork-postponed-until-9-20-2016
Kinda feel like this post might have made a small difference.