How Much Do Whales Influences on Rewards? Introducing New Statistics

clayop (73) in stats • 8 years ago (edited)

Summary

This post introduces two new statistics about influences on rewards with the consideration of the leverage of Steem Power (IPR) to provide more through understanding about the whale issue.

Introduction

The whale issue of Steem is one of the most frequently debated topic from its beginning. But we still do not have through understanding regarding the current situations, such as how much whales have influences on rewards, or how unequal the situation is. In this post, I would like to introduce two new statistics, 1) Influences on (author) rewards, 2) Influence to Steem Power Ratio (IPR), to provide more comprehensive knowledge about the whale issue.

Methods

The data is from Steem blockchain from 2016/8/16 (After v0.13.0) to 2016/9/8.
Influences of whales on rewards is calculated based on the following equations.

- Rshares of a post: sum(voters_rshares)
- Vshares of a post: (posts_rshares)^2
- Daily vshares: sum(posts_vshares)
- Voter's vshares on a post: posts_vshares * (voters_rshares / posts_rshares)
- Voter's daily vshares: sum(voters_vshares)
- Voter's influence: voters_daily_vshares / daily_vshares

To get IPR, we just need a simple equation.

- IPR = Influence / (voters_vests / total_vests)

Note: Steemit account's vests are excluded from total_vests

Results

Rank	Account	Influence(%)	Steem Power (%)	IPR
1	smooth	13.34	2.86	4.67
2	blocktrades	11.60	3.71	3.13
3	berniesanders	10.55	2.97	3.55
4	steemed	3.92	1.62	2.42
5	itsascam	3.82	1.53	2.49
6	tombstone	3.62	1.87	1.93
7	dantheman	3.51	2.54	1.38
8	summon	3.11	1.29	2.41
9	jamesc	2.78	3.20	0.87
10	complexring	2.37	0.67	3.52
11	smooth.witness	2.32	0.53	4.37
12	wang	1.97	0.66	2.97
13	nextgencrypto	1.81	0.51	3.53
14	rainman	1.78	1.51	1.18
15	recursive	1.56	0.31	5.11
16	silversteem	1.55	0.46	3.39
17	riverhead	1.53	0.68	2.23
18	pharesim	1.53	0.88	1.74
19	ned	1.52	5.73	0.27
20	steemit200	1.38	0.57	2.42
21	steempty	1.30	0.69	1.87
22	kushed	1.15	0.50	2.29
23	xeldal	1.15	0.64	1.79
24	hr1	1.00	0.21	4.69
25	satoshifund	0.99	0.35	2.85
26	enki	0.91	0.53	1.71
27	val-a	0.91	2.80	0.32
28	silver	0.74	0.19	3.85
29	au1nethyb1	0.72	0.42	1.71
30	badassmother	0.63	0.26	2.44

Surprisingly, top 10 have about 60% (actually over 60% since smooth and smooth.witness is controlled by same person) of author rewards, while they only have 22% of Steem Power. And the top 3 (smooth, blocktrades, berniesanders consist of 35.5% of influences, for instance, if daily author reward is $20,000, about $7,000 is from these three's voting activities. The average IPR of the top 10 is 2.63, but if we do the same calculation with rank 41~50, they only have IPR of 0.75. This means that influences on rewards is unequally distributed, more specifically it is biased to some whales.

Additionally, I calculated some basic statics about top rankers voting patterns.

Range: Number of unique writers they voted for
Mean: Average number of votes per writer
Max: Maximum number of votes on a writer
Stdev: Standard deviation of votes

Rank	Account	Range	Mean	Max	Stdev
1	smooth	340	2.32	65	7.49
2	blocktrades	480	1.31	22.4	1.51
3	berniesanders	785	1.27	14.5	1.7
4	steemed	80	3.64	44	6.43
5	itsascam	79	3.69	44	6.49
6	tombstone	196	1.64	12.1	2.18
7	dantheman	339	0.66	10.3	1.23
8	summon	220	1.65	14.4	1.99
9	jamesc	135	1.1	12.8	1.81
10	complexring	534	1.75	39	3.04
11	smooth.witness	324	2.4	65	7.61
12	wang	213	8.29	82	10.41
13	nextgencrypto	755	1.4	14	1.51
14	rainman	170	1.18	9.7	1.3
15	recursive	121	9.78	51	11.76
16	silversteem	760	1.82	15	2.08
17	riverhead	334	1.53	12.6	1.4
18	pharesim	404	2.03	54	3.89
19	ned	238	0.18	2.6	0.35
20	steemit200	175	2.41	13	2.29
21	steempty	164	2.89	34	4.84
22	kushed	250	2.61	84	7.36
23	xeldal	204	2.92	48	5.65
24	hr1	104	10.34	52	11.86
25	satoshifund	130	2.51	21	3.61
26	enki	174	3.22	46	5.58
27	val-a	75	0.36	1.5	0.42
28	silver	763	1.83	15	2.13
29	au1nethyb1	205	1.97	11	1.79
30	badassmother	235	2.75	15	2.77

Among these statistics, max and stdev should be highlighted, since higher numbers imply that a whale's voting is very concentrated (possibily due to his/her preferences or favorable connection with authors?), and this can further create perception of inequality. (People recursively see high rewards on same author's posts). For comparison, while blocktrades and berniesanders have similar influences with smooth, they have much lower max and stdev, which implies that their votes are widely dispersed to many writers. But we can see very high max and stdev in hr1, recursive, wang, and smooth. In addition, we can find smooth and smooth.witness, and recursive and hr1 have the similar patterns respectively, probably because each pair of accounts is using the same voting bot.

Wrap-up

The results show that influence is much higher than whales' actual steem power, especially by active whale bots. In my humble opinion, the fundamental cause is the squared rshares, which makes large voting power much greater. We need further discussion about this issue and have to address it for better Steem society.

P.S. Do you think that I have to post these stats everyday, weekly, or monthly? (Everyday data has influence and IPR only because the second table requires more samples)

stats steemstats steem steemit

8 years ago by clayop (73)

$755.46

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!

Sort Order:

Trending

[-]

smooth (70) · 8 years ago (edited)

@clayop I'm frankly disappointed to see this kind of divisive and inflammatory rhetoric from you.

The reason my votes or others high on your list have the more relative influence (including relative to other whales with the same or higher SP, and therefore the same exact weighting given the mathematical formulas behind curation) is that we have chosen to invest our time and/or resources on actually curating. In my case I hired a team which reviewed posts on the site for many person-hours every single day, as well as spending a lot of my own time on it. Blocktrades I believe has his wife spending a lot of hours curating (I don't know if it works out to "full time" or what but it certainly seems like it could).

By contrast, other whales such as yourself, have chosen to spend relatively little to no time on curating, and thus have relatively little influence on content or rewards. This is not because you are being shut out by the system, it is because you chose not to participate. Given your SP relative to others on the list (such as @kushed or @steempty for example, who have approximately the same SP), you certainly could have had influence if you had made the investment.

SP does not give automatic influence or rewards. It gives the opportunity to exercise this influence by making an investment in curation. It is up to the SP holder whether or not to do so.

BTW, one clarification:

In addition, we can find smooth and smooth.witness, and recursive and hr1 have the similar patterns respectively, probably because each pair of accounts is using the same voting bot.

Your analysis is not correct here. The @smooth.witness account is entirely run by a bot which mirrors every single vote made by @smooth (including downvotes, low power votes, removal of votes, etc.). Any differences between the two are due to either malfunctions or very rare occasions when I have manually overridden this mirroring (for example because and author asked me to reduce voting on a post and the easiest way to do this was by removing only one of the votes).

This is done because I own both accounts despite them being separate on the blockchain and treat both as being my own for voting purposes (unlike other such as yourself, blocktrades, etc. who use their primary account for their witness). Both of these accounts are not run by the same voting bot, in any manner whatsoever.

$1.02

[-]

jamesc (61) · 8 years ago

I'm happy smooth is what we like to call awake!! I have not checked all your votes but I have seen you make some fantastic votes in terms of help other people learn some vital information for safety and healthy.

$0.34

12 votes

[-]

clayop (73) · 8 years ago

I agree your mirroring bot can cause the result.
However, in terms of curation, @blocktrades and you have obvious differences. Your votes is concentrated on few writers, more specifically, top 5%(17 people out of 340) consists of 65% of your vote (528 votes out of 816). For instance, knozaki2015 got 65, terrycraft 53, crazymumzysa, gavvet, sirwinchester got over 40. Meanwhile, @blocktrades 's top 5% (24 out of 480) has 20.6% only, which is three times lower than yours. Over 10 votes on the same author in the study period is 3, while @smooth has 19.

P.S. you even did't vote for this stats while you always vote for masteryodas daily payout. :)

$0.13

5 votes

[-]

bitcalm (64) · 8 years ago

Can we conclude that the reason the trending page always has the same damn authors is because @smooth's votes are limited to fewer authors?

I think whales should take a lesson from @robinhoodwhale and not upvote the same author in a given time period.

I for one am tired of seeing the same set of authors on the trending page, each post earning almost exactly what their last post earned.

$0.06

2 votes

[-]

clayop (73) · 8 years ago

Actually, the authors smooth repeatedly voted are not on the trending page often. But they got significant amount of author rewards during the last month (between $200 and $400 generally)
Please see the following accounts: @knozaki2015 @terrycraft @crazymumzysa @gavvet @sirwinchester @masteryoda

$0.00

[-]

bitcalm (64) · 8 years ago

I'm surprised by that to be honest. I wonder what is causing the same authors to appear on the trending page.

$0.00

[-]

logic (67) · 8 years ago

From my experience, berniesanders, phareism and block trades have been the most generous whales when it comes to regular upvoting of minnows posts.

$0.00

[-]

smooth (70) · 8 years ago (edited)

However, in terms of curation, @blocktrades and you have obvious differences

Sure, people have different strategies for voting. @blocktrades also has way more votes called out for abuse (e.g. plagiarism, very questionable value content, etc.) than I do. I'm not even saying that is a bad thing, it is kind of a trade-off with voting on a lot of unknown stuff. (I do vote on unknowns too but they are scrutinized a lot more heavily with I'm reasonable sure more time put into vetting.)

P.S. you even did't vote for this stats while you always vote for masteryodas daily payout. :)

Perhaps a valid point (though no I don't always vote for his reports).

There is a huge difference. His reports are usually presented in a neutral data-speaks-for-itself manner. Even when he editorializes he avoids divisiveness and criticism. As such, his posts generally attract positive and good-vibe comments that add value to the community and make people feel good about being here. That is how you retain and attract a user base.

Your post, while the statistical analysis is valid and I recognize the work, is presented in a manner that is divisive and feeds on (as well as feeds) jealousy and hostility, and greatly oversimplifies the reasons for differences in (what you call) influence between users, as I pointed out above. While I think your goals in wanting Steem to succeed are sincere and I share them, your approach is misguided and instead spreads a negative vibe, drives people away, and does not attract new users. I do not feel that adds value or helps to make Steem a better platform.

$0.00

[-]

clayop (73) · 8 years ago (edited)

You're labeling critical interpretation as jealousy and hostility IMO. I think you maybe not comfortable since you are the on the top of the issue.

$0.00

[-]

hannixx42 (59) · 8 years ago

I get what you are saying @smooth. But my question is how would you view my investment being a dedicated human curator who votes daily for his 106 day straight? My account value is what a full time curator would earn botless. What we need is a solid base of human curators and bots in my opinion are taking these funds right out from under the human curators only to distribute them to authors they privy. We need your help to grow curators which in turn give more value to authors. Giving to authors is great but we need a solid base of voters to promote rewards that grow steadily. $100-$1000 articles are killing us. We need rewards like $1-50 so that the value of rewards can grow in time and be more distributed to more authors.

$0.83

4 votes

[-]

clayop (73) · 8 years ago

@furion I have another thoughts on your result. High profitability in curation probably means that a curator voted for popular posts, and then he/she can maximize the curation rewards. Meanwhile the low profitability can have two cases. 1) low profitability with wide range of votes (in my data, low max and stdev). 2) low profitability with small range of votes (high max and stdev). The first case implies that a curator is a hidden gem finder. He/she wants to find new and good authors. The second case implies that the curator is a sub-community supporter. The curator's vote is loyal to small group of people in his/her community. I think smooth is this case. But I am not sure what kind of community he is supporting now.

$0.05

1 vote

[-]

clayop (73) · 8 years ago

As you told, more dispersed voting does not necessarily mean better curation. However, concentrated voting is more likely to be a barrier against new users. For instance, @knozaki2015 has lots of over $300 posts for a month (probably after you started voting for him?), which is almost impossible to average users. IMHO, you can reduce your max by decreasing upvote percentage and still keep support writers.

$0.00

1 vote

[-]

smooth (70) · 8 years ago (edited)

The concentrated vote implies less effort finding hidden gems

@clayop voting for "more authors" does not necessarily mean doing the best job of curation, nor does it imply "more effort". I'm accusing no one of anything but one could easy create an army of sock puppets and vote for each of their posts, creating an appearance of "voting for many different authors", but it would all be fake and self-serving. Perhaps slightly less sinister but also useless, one could simply browse New and vote on any random crap posts, producing a small-but-consistent reward with minimal effort and in doing so also vote up a large number of different authors.

I'm happy that the authors I've consistently supported have produced content that gets votes from other community members, sparks engaged conversations in the comment sections, and is downvoted rarely. I've also voted on a good number hidden gems that then went on to get many other votes after I discovered them. Those are the primary metrics I use, along with my subjective view on what is adding value to the platform and generally helping it to thrive in various ways.

$0.00

1 vote

[-]

furion (70) · 8 years ago

@clayop check out the report I've released yesterday on top curators by profitability. @smooth is not even on the list, because his curation profitability is ridiculously low (7% PA). On the opposite side of the spectrum we have for-profit whale bots, that generate 30%+ returns.

If smooth was selfish, or malicious, this probably wouldn't be the case. As far as I know, he is the only person that spends a lot of time and money on curation that is not based on ROI - rather I believe its his genuine intent to provide value to the platform and use his voting rights to support quality content.

$0.00

1 vote

[-]

smooth (70) · 8 years ago

I'm not uncomfortable with my results, I consider them an decent accomplishment (but I can always strive to do better).

I disagree with you making me, or anyone else, being at the top of the list into an "issue" at all. I've invested the most in terms of a combination of SP and effort and that is why I'm at the top. You've invested apparently little or no effort so you aren't on the list at all, despite people with less SP than you being there. That is exactly how it is supposed to work.

$0.00

[-]

clayop (73) · 8 years ago

Yes first table is representing the drgree of utilization, and you have done a good job. But second table shows how you utilize. The concentrated vote implies less effort finding hidden gems, instead a voter strongly support only few people.

$0.00

[-]

clayop (73) · 8 years ago

@furion First of all, this statistics is about author rewards, not curation rewards. Secondly, this statistics does not tell @smooth is selfish or malicious. The results present that smooth is the most influential person on author rewards and his votes are concentrated on small group of people.
About your results, probably bots can maximize their curation rewards by voting on high reward posts.

$0.00

[-]

abit (70) · 8 years ago

The second case implies that the curator is a sub-community supporter. The curator's vote is loyal to small group of people in his/her community.

IMO that's an effect of the follow feature.

$0.00

[-]

fast2future (50) · 8 years ago

Everybody, keep on keepin' on - one day at a time to Steem greatness!!!

$0.00

[-]

ltm (57) · 8 years ago

The botting world is so fascinating, wish i knew more!

$0.00

[-]

morning (70) · 8 years ago (edited)

Great posting. Many of us were feeling unfairness, while having no data to support the opinion. Now I can say whales Influence is way too large compared to their Steem Power.

$0.78

3 votes

[-]

publicworker (59) · 8 years ago

Great analysis sir!

$0.67

1 vote

[-]

krnel (78) · 8 years ago

Very useful to see how the power is distributed. How did it get this way? First come first serve means your position is set for life? And that's all there is to it?

$0.07

2 votes

[-]

patelincho (67) · 8 years ago

Funny you get 160 plus upvote ( at the moment i look ) and you will be paied less than a single 7 upvotes in some posts

$0.06

2 votes

[-]

timcliff (73) · 8 years ago

We need further discussion about this issue and have to address it for better Steem society.

Curious your thoughts on this:
Human vs. Bot Curators - Introducing Human Competition into the Equation - Allow Whales to Delegate Curation Power to Dolphins and Minnows

$0.00

1 vote

[-]

hannixx42 (59) · 8 years ago

And this: Bot War: Keeping Peace with a Reverse Fractional Formula

$0.00

[-]

jovic (30) · 8 years ago

The statistics are shocking. I think that the whales have too much power and too much influence. It's time the minnows are given more say

$0.00

1 vote

[-]

clayop (73) · 8 years ago

It would be interesting to trace changes after the next hardfork with vote balancing
Rebalanced target votes per day from 40 to 5. Votes will still regenerate over a 5 day period, so an account can vote 25 times sporadically before incurring large penalties to voting power, or 5 times daily.

Currently, top influencing whales voted much more than 5 time a day (e.g. smooth did 33 times on average, and wang did 73 times!) The change will dramatically alleviate the concentration of influences.

$0.00

[-]

cryptos (66) · 8 years ago (edited)

Regular users do feel that most Whales are not doing much good in utilizing their serious voting power and now it can be seen in numbers as well thanks to your stats...

I have been doing some Whale Watching on my own lately, tracking the votes of something like 60 of the top Whales that actually curate posts and what I'm seeing is something like 4:1 of ratio of Whale votes going for popular and established authors compared to new users.

$0.00

3 votes

[-]

alexgr (66) · 8 years ago

Are slidebar percentages calculated? I mean if a curator has given 50 votes x 10% to an author they prefer, it's like 5 votes (or less) x 100% voting power.

Yet the first case would seem "bad" because it would appear there are too many votes on the same people.

$0.00

1 vote

[-]

clayop (73) · 8 years ago

Yes it is considered.

$0.00

[-]

whatsup (75) · 8 years ago

Great information and it will be interesting to see who votes on this! :0

$0.00

1 vote

[-]

chitty (69) · 8 years ago

I can´t upvote this enough, those are some great stats that explain how biased the rewards are towards large holders. Do you think this situation will improve after the hardfork on tuesday?

$0.00

[-]

complexring (68) · 8 years ago

That's what superlinearity does ... and it's necessary to prevent sybil. Even if it were minutely superlinear, in the long run, that will ALWAYS favor large stake holders. There's no way escaping that math.

$0.00

1 vote

[-]

mrweed (57) · 8 years ago

"We need further discussion about this issue and have to address it for better Steem society."

I believe this is key in order for Steem to get back on track.

Great input !

thanks

$0.00

[-]

littlescribe (65) · 8 years ago (edited)

This is great! Thanks for sharing! I featured your stats in my article talking about how a SHARE button would decentralize some of this: https://steemit.com/share/@littlescribe/a-share-button-is-it-really-a-good-idea

$0.00

[-]

thecryptofiend (74) · 8 years ago

Very interesting analysis. Thank you:)

$0.00

[-]

vi1son (59) · 8 years ago

How can there be Max a fractional number? And Mean below 1? Are you count % of power used for that?

$0.00

[-]

vonpetro (55) · 8 years ago

in the long term, as more topics and keywords are generated, and more people get involved with steem, wont that smooth the voting power for all the people posting who don't receive whale love? in essence wont the longer steem is around the better the voting, the better the payouts without whales.

$0.00

[-]

steempowerwhale (59) · 8 years ago

thanks so much for this statistic. I am fascinated by the blockchain and what data analysis it enables. Would you know which of these whales are human voters and which ones are bots?

@steempowerwhale 🐳
🌞 upvoting your lifetime dreams!

$0.00

[-]

clayop (73) · 8 years ago

I cannot recognize them accurately, but high stdev may give some clues.

$0.00

[-]

cryptomental (59) · 8 years ago (edited)

One way to recognize whale BOTs would be to get timestamps of their votes and calculate the delta from the time that the post was submitted. I guess you would find a few interesting heuristics there. This is also a way to partially reverse engineer BOTs taking part in the bot war.

$0.00

[-]

abit (70) · 8 years ago

Almost agree, however whales can hire people in different timezone for curating 24 hours a day.

$0.00

[-]

jamesc (61) · 8 years ago

Max: Maximum number of votes on a writer

Interesting that you have a decimal place. Is this because your considering the vote weight?

$0.00

[-]

clayop (73) · 8 years ago

Exactly

$0.00

[-]

good-karma (76) · 8 years ago (edited)

It gives different angle and much more clear stats I have seen so far. Good work on stats, though of course whales have preferences and favored authors. And would it be possible to make same stats for last week or two weeks ago data? Maybe we would know if anything changed, what actions can be taken? I think it would give insights to whales to see if their actions/ways are working/improved.

$0.00

[-]

dennygalindo (59) · 8 years ago

I think reducing the power exponent (from 2 to 1.75 or 1.5) and reducing number of whale votes would greatly help reward distribution

$0.00

[-]

lemooljiang (73) · 8 years ago

The best solution is to Decentralized power.

$0.00

[-]

jamesc (61) · 8 years ago (edited)

Help me out here. If a person votes for a $1 post they may influence payout at 1% but if the same vote goes for a $10 post the payout is adjusted at 10%. Is this represented in your calculations above? Please forgive me, I did not crunch the numbers ;)...

$0.00

[-]

clayop (73) · 8 years ago

Ok let me give you an example.
Case 1: Total daily vshares = 100M, a post's rshares = 1,000 then a post's vshares = 1,000^2 = 1M (1% of the total payout). If a person gives 100 rshares to the post, he will get 1M * 100/1000 = 100,000 vshares (0.1% of the total vshares)
Case 2: Total daily vshares = 10M, a post's rshares = 1,000 then a post's vshares = 1,000^2 = 1M (10% of the total payout). If a person gives 100 rshares to the post, he will get 1M * 100/1000 = 100,000 vshares (1% of the total vshares)

$0.00

1 vote

[-]

slowwalker (80) · 8 years ago

The thing is that unfairness exists in steemit.
I fully agree with @clayop.
Problem must be fixed.
Steemit is not for some selfish ones but for ordinary peaple who want to exchange information and knowledge.
Economic reward should be secondary purpose.
If not, Steemit will wane down.

$0.00

[-]

linkback-bot-v0 (54) · 8 years ago

This post has been linked to from another place on Steem.

Learn more about linkback bot v0.3

Upvote if you want the bot to continue posting linkbacks for your posts. Flag if otherwise. Built by @ontofractal

$0.00