This article was orginally published on www.cryptogo.de
Everyone is talking about Delegated Proof-of-Stake (DPoS) due to the upcoming mainnet launch of EOS. Besides the EOS DPoS family, which BitShares and Steem are a part of, there is also the far larger Lisk-DPoS family. That especially the Lisk governance system looks more like the Italian Mafia by now than a decentralized network only very few outside of the DPoS community know. For this reason this article will clearly name the problems leading to mafia like structures and will underpin these problems with facts before pointing out possible solutions in the end.
How does Delegated Proof-of-Stake (DPoS) work?
Delegated Proof-of-Stake maps the american electoral system onto the blockchain. Tokenholders elect delegates. It is their task to confirm transactions and forge blocks. Here the term “forging” can be roughly equalled to the term “mining”. For their work the delegates receive a reward per block (forging rewards) as well as the fees contained in the block’s transactions. Often, the delegate shares his profits with his voters through so-called “kickback payments”. To run for the delegate elections only a one time fee of currently 25 Lisk has to be paid. Through every election in the Lisk network 101 active delegates are chosen.
What is the incentive for the delegates?
In the Lisk network the maximal possible amount of forging rewards per year for every delegates comes to about 1,5 million US dollar (Status May 2018). Due to the high revenues and low entry threshold it’s no wonder the delegates are playing hardball for the favour of the voters.
Thus, over the course of time, cartels have formed with the goal of establishing the same group of delegates in the top 101 as permanently as possible. Some problematic features of the Lisk network are playing into these cartels hands here.
Which problems exist within Lisk?
- low voter turnout
- voting is too expensive for users with few tokens
- kickback payments offer little incentive to participate in the election
- flawed incentive system consolidates the power of delegates
- lacking fluctuation of delegates leads to formation of an oligopoly
- cartels consolidate the power of delegates
1. low voter turnout
In Lisk every token holder can use their voting power to vote for for a maximum of 101 delegates. If a user holds 1000 Lisk their votes have a voting weight of 1000. It doesn’t matter if they cast a single vote or all 101 of them their voting weight is always 1000 for every single vote cast.
In this system a high voter turnout is very important. In the case of a low turnout whales, meaning accounts with a (very) large amount tokens, have a larger impact on the outcome due to their voting weight. If a whale owns 1% of all tokens with a voter turnout of 100% he has exactly 1% of the total voting weight. The lower the accumulated voting weight of votes cast is the higher the influence of the whale’s voting weight is. Because of this their vote is getting more and more important for every single delegate. A low voter turnout can therefore grant a whale a lot of influence and with that a lot of power in the network.
The chart above shows statistics about core parameters of the network. It should be particularly surprising that only 21% of active accounts, so accounts with a balance of larger zero, are using their right to vote. But these 21% in turn have 42% of the total voting power.
That 42,82% of all Lisk is held by only 10 accounts might be a red flag for some.
Here let it be said that on the one hand most coins of the Lisk family, i.e. Ark or OxyCoin, exhibit a higher value between 60% and 70% degrees of centralization. On the other hand this centralization of tokens mostly is a result of Wallets of the Lisk team or Hot Wallets of large exchanges.
However, of these 10 accounts only Oliver Beddows, one of the founders of Lisk, is making use of his right to vote so that these accounts aren’t wielding all that much influence.
2. voting is too expensive for users with few tokens
The chart below offers an overview of the distribution of wealth inside the Lisk network. Barely 45% of all accounts hold a maximum of 10 Lisk and a further 17% of accounts hold between 10 and 50 Lisk. Therefore 62% of all accounts hold a maximum of 50 Lisk.
But these numbers only gain real meaning when put in relation to the arising fees for voting in the election of delegates. For 1 Lisk 33 votes can be cast. The maximum number of 101 votes costs 4 Lisk. This makes it clear that almost half of the owners of Lisk tokens would also loose half of their tokens by participating in the election. If you remember that in this category not every account holds 10 Lisk but accounts holding up to 10 Lisk are aggregated the statement above is rather optimistic.
These high fees make it doubtful that small investors will or even can make us of their right to vote. This in turn would mean that about 45% of the voting weight are precluded from the election.
The following chart shows a statistic about the relation between votes and voting power. It makes it very clear that the accounts with a capital of more than 10.000 Lisk are only responsible for 6% of all active votes but these votes command more than 83% of the voting weight. In contrast voters with an account balance of up to 10 Lisk come to 23% of all votes but don’t really play any significant part with only 0,01% of the total voting power.
In general the number of “real” votes by voters with up to 10 Lisk can only be speculated. The large number of votes can probably be attributed to “old voters”. These sold all their Lisk but a tiny amount after casting their votes. This thesis is supported by the fact that 55% of this category only own a maximum of 1 Lisk.
To properly sort these numbers into the overall context one best asks the question: How much voting power does a delegate need to be voted into the top 101 of delegates? At the current point in time this needs about 26–27% of the total possible voting weight. If you base this on the sum of the voting power of all voters who are actually using their right to vote you would need 62% of the voting power.
3. kickback payments offer little incentive to participate in the election
To get token holders from the virtual couch to the virtual ballot box many DPoS networks allow the distribution of revenues to their voters. Lisk, too, allows such kickback payments. Every delegate can freely decide how many percents of their revenue they want to share with their voters. Through kickback payments the voters can regain their voting fees as well as generate additional profits which might also move small investors into casting their votes.
But if you look at the maximum kickback revenues based on optimized voting those small investors mentioned will quickly sober up. A voter with an account balance of 10 Lisk will take about 24 weeks after deduction of the full fees to regain the initial balance of 10 Lisk. This includes compound interest. With a capital of 25 Lisk it only takes 8 weeks, with 50 Lisk it takes only 4 weeks to completely regain the fees.
So if these kickback payments are an incentive for small investors to make use of their right to vote is questionable. Should potential voters not be able to wait these 8 respectively 4 weeks for their investment in the worst case 54% respectively 62% of all voters would not step up to the ballot box.
A look at the statistical distribution of nonvoters supports this apprehension, too. Just about 10% of all token holders with a maximum of 100 Lisk make use of their right to vote. Since the already mentioned “old voters” only own a small absolute sum of lisk their influence on the validity of this statistic is only marginal.
4. flawed incentive system consolidates the power of delegates
A further problem is the distribution of the forging rewards among the delegates. As already mentioned the 101 delegates with the most votes receive the opportunity to forge blocks and collect rewards. To be able to keep counting on their votes in most cases a freely selectable percentage of these rewards is shared with their voters by the delegates.
By implication the voters will for economical reasons only vote for delegates who have a high chance of participating in the production of blocks again. If the vote is given to a delegate outside of the top 101 the voter doesn’t receive kickback payments. Through this incentive system the reelection of already active delegates is highly favored. The result is a high barrier of entry for delegates outside of the top 101 and therefore a very low fluctuation among the active delegates. The power inside the network is consequently spread onto only a few shoulders.
This theory can also tellingly be proven by a chart about the absolute number of votes per delegate rank. Even without a key you could tell the border between active and passive delegates.
5. lacking fluctuation of delegates leads to formation of an oligopoly
Up until now we have had a look at the general voting system as well as the incentives for voters. Now we will have a more detailed look at the already mentioned economical incentives for the delegates.
After every election every elected delegate can extend the blockchain by one block and for this receives the transaction fees accrued in their block as well as currently a reward of 4 Lisk. With the time between two blocks defined by the system to 10 seconds it takes about 17 minutes (1010 seconds) until all 101 elected delegates can produce their block in a single round. So every one of these delegates has the opportunity to produce a block 85 times a day ((24h*60m*60s)/1010 seconds). So only through forging rewards they earn about 340 Lisk (85 blocks * 4 Lisk) a day. With a current price of 12$ a delegate who is continously voted into the top 101 earns 4080$ a day, 124.100$ a month and in a year almost 1,5 million US-Dollar (as of May 2018). Remember that these earnings are generated through forging rewards alone. Additionally collected transaction fees are not yet considered in this calculation.
But these barely carry weight in comparison to the block rewards. The chart below illustrates the development of revenues per delegate and month from transaction fees. Since the existence of the Lisk network the monthly revenue through these fees was far below the daily revenue through block rewards.
Even if part of the revenues are distributed among voters the system still allows for a very lucrative income.
As we already know the fluctuation of active delegates is not exactly high due to the offered incentives for voters. The combination of a low fluctuation and a lucrative income opportunity inevitably leads to the formation of an oligopoly.
6. Cartels consolidate the power
After the numbers presented in previous paragraphs it should be no surprise that all active delegates have a strong incentive to be able to keep producing blocks. To improve the odds for this at first several alliances had been forged with the goal to support each other with votes. In the meantime these alliances have grown to full blown cartels in almost all Lisk-Fork based DPoS networks. However these agreements are not arranged in dark back-rooms but are done publicly through pressure exerted on the voters. Thus the largest Lisk cartel “Elite” demands that a voter uses all their votes for delegates of the Elite cartel to receive kickback payments. The second largest cartel “GDT” acts similarly. Here, too, the users have to vote for all the members of the cartel to receive kickback payments in full. This is strongly reminiscent of the methods of mafia-like clans and not of a decentralized network.
To completely understand the scale of these cartels a look at their size is enough. “Elite” provides 54 of all 101 delegates at the time of the snapshot and “GDT” 32 delegates. This means that 85% of the entire block production has been taken over by two cartels. The only bright spots are the Lisk ecosystem funds in which “Elite” pays 5% and “GDT” 10% of their earnings. This money is used to further developments all around the Lisk network.
The registration of new delegates is at a historic low, too. If these are consequences of the consolidated balance of power or are due to other reasons could only be speculated here.
Potential problem-solving approaches
The essentials of the problems worked out in this article have been known inside the DPoS community for a while now. The statistical analysis and the graphic editing now give the rest of the community a clear approach to this topic.
Even the team behind Lisk had to admit to some problems in their network in February 2017. The prospect of some changes had been presented back then:
- Reduce the number of votes to 33 per account
- Change the formula for the calculation of voting weight:
“(Amount of Lisk * 33) / number of votes cast”
- Dynamic transaction fees to lower the costs of all transactions
- Changed distribution of block rewards to also reward the willingness to serve of delegates outside the top 101. This change is supposed to break the centralized power structure of the delegates and lead to a greater fluctuation inside the top 101
None of these meaningful changes have been realized up until now. In the context of the Consensus conference 2018 the Lisk founder Oliver Beddows assured CryptoGo that especially the dynamic transaction fees are supposed to be introduced this year. After this change the fee structure would no longer be static but would orient itself at the size of the transaction, similar to Bitcoin. This is certainly a sensible step. But if and how this change can contribute to the solving of the before mentioned problems waits to be seen.
Another possible solution is the complete ban of kickback payments. EOS is the first larger DPoS network to approach this subject. But it is questionable if indirect kickback payments, i.e. through tangible assets or payments in other currencies, even can be prevented.
Other Lisk-Forks have already experimented with changed parameters of the consent algorithm. But here, too, signs of heavy centralization and cartel formation showed.
Our analysis has shown that DPoS networks can be very fragile constructs. In this the incentives between participating parties need to be balanced and adjusted, if necessary, to reach as high a degree of decentralization as possible. According to the opinion of CryptoGo this degree is essential for the long term success of every network.
Lisk shows impressively what happens if a network gets off balance. Even if the solutions listed above were implemented it would remain questionable if the delegates would accept these changes and therefore rob themselves of part of their revenue.
Thanks to hasufly for his feedback!
Should you have further questions, found mistakes within the review or would just like to discuss the newest developments in the cryptoworld with the CryptoGo-team then become part of the CryptoGo community and join our Telegram-Channel.
I am always happy about feedback: simon[at]cryptogo.de
CryptoGo has the aspiration to transparently and independently report on and evaluate new ICOs. In doing so we create the articles to the best of our knowledge and belief. But the reports and evaluations on this site only reflect our own opinion and in no way are meant to issue investment advice. Every investor should extensively investigate the presented ICOs and analyze and asses them to their own standards.