Author | Binance Research

• Due to EOS's high market cap, as a PoS consensus network, it is often singled out and considered a victim of its governance. The largest EOS holders among them have further cemented their position and wielded all the power.

• The decentralization of EOS governance can be well assessed by looking at the proxies of EOS.

• The evaluation was conducted in three ways, looking at the performance of EOS in terms of collusion resistance, fault tolerance, and attack resistance.

• There are the following insights about collusion resistance:

1. EOS governance lacks mechanisms to avoid or structure voting transaction processes

2. EOS’ Governance Enforces Mergers and Even Ends Up Vote Trading and Selfish Behavior

3. Individual institutions such as proxies or block.one have sufficient power to influence votes.

• Regarding fault tolerance:

1. Operational metrics measuring reliability and responsiveness show that 2/3 of the 21 block producers (BPs) performed poorly.

2. In addition, two failures occurred.

• In general, EOS problems seem to be caused and exacerbated by many issues, such as low turnout, poor resistance to Sybil attacks, poor consistency transparency, mechanism where 1 EOS can vote 30, and changed blocks award.

• Finally, in terms of attack resistance, there appear to be two block producer clusters, evident in the correlation between voting patterns and regional distribution.

“If a startup’s potential is directly proportional to the capabilities of its competitors, the most promising startups will be the ones competing with governments. Don’t think that’s impossible. That’s what digital currencies are for.”

—— Paul Graham, Y-combinator

secondary title

Introduction to EOS and EOS Governance

1.1 Introduction to EOS

Block.one raised over $4.1 billion in its one-year token funding as of July 2017. This is the largest token funding round on record, surpassing the next largest token funding round ($257 million for Filecoin) by nearly 16x. Subsequently, Block.one officially launched EOS on the mainnet on June 1, 2018.

image description

Data source: CoinMarketCap, Binance Research

Not only the price, but also the EOS ecosystem is growing rapidly. EOS is now one of the most popular blockchains for dApp development. At the time of writing, data from Dapp.Review shows that 676 dApps have chosen to build in EOS, a number second only to Tron (693 dApps) and Ethereum (2195 dApps).

These EOS dApps are also being used. In fact, their activity far exceeds that of dApps on Tron or Ethereum. At the time of writing, Dapp.Review recorded that almost 99% of transactions (nearly 32 million) involved dApp smart contracts on EOS. The respective transaction volumes are an order of magnitude higher than dApps on Tron (approximately 330,000 transactions), which is still more than double the number of dApp transactions on Ethereum (approximately 150,000 transactions).

This observation is even more startling when considering the number of users per chain. Three months ago (October 30, 2019), the number of EOS users was close to 85,000, but now the number of EOS users (about 10,000) is the least among the three blockchains. Ethereum clearly leads the pack with the largest number of users (around 50,000), more than double the number of Tron users (around 24,000).

Proxy measures of such activity (transactions involving dApp smart contracts) can be interpreted in a variety of ways and can provide insights into:

• The complexity of dApps, as more complex dApps may utilize multiple smart contracts.

• For example, dApps like games may require more frequent user interaction.

Most relevant to the interpretation of this indicator, however, is the underlying technical infrastructure. Unlike Ethereum (still) which uses a PoW consensus mechanism, EOS uses a dPoS mechanism. Although dPoS can improve network throughput, because it is based on the "institutional reputation" of a small number of participants, it comes at the cost of decentralization.

The EOS infrastructure uses a group of 21 nodes (also known as supernodes) that can vote on new blocks in a round-robin model. These nodes are selected by EOS token holders from a large pool of candidate BPs. Since block producers are rewarded at every block validation, they have an incentive to be elected as block producers, which puts them in direct competition with each other for votes. Block rewards are paid through annual token inflation.

EOS token holders execute votes for specific block producers by staking tokens within 3 days. Over time, votes "decay" and become invalid after two years. To maintain high voting intensity, votes must be resubmitted weekly. While token ownership and voting power typically scale linearly at a 1:1 ratio, it is also possible to vote for up to 30 block producers simultaneously. In practice, this means that one token may equal 30 votes.

In addition to voting rights, token holders typically receive a prorated block reward for the rewards their block producers earn, and may similarly gain access to network resources such as RAM, CPU, and bandwidth.

To avoid inefficient allocation of resources by inactive token holders, additional market-driven allocation mechanisms need to be introduced.

1.2 Is EOS too centralized?

While this particular problem of optimizing resource utilization on the EOS network is easily overcome by stakeholder-driven innovation factors, other problems become more persistent. In particular, concerns that EOS may be too centralized have accompanied EOS from its initial token financing to its present state, and WeissCrypto talked about the reason in 2019 because "too few people own too many tokens."

In general, all blockchains are touted as:

“Models prone to redecentralization: they are informally controlled by powerful actors in the digital currency ecosystem who may violate the basic rules of the blockchain community without accountability or sanction”.

To answer the question that EOS may be too centralized, the opaque notion of centralization must first be defined.

1.3 Measuring decentralization

The definition and measurement of "centralization" has long been in the spotlight. Especially in the early days of BTC and digital asset adoption, "decentralization" has been a guiding goal.

But in 2017, Buterin’s blog post sparked a discussion that became less dogmatic and more pragmatic in terms of the purpose and benefits of decentralization. Until then, people think that there must be a move towards decentralization. Also, this concept was well represented in the early debates about BTC block sizes. Most of the discourse is not framed in terms of "value of decentralization" or cost-effectiveness, but in absolute ideology.

Butterin builds on previously existing research, such as the popular opinion paper "On Scaling decentralized Blockchains" (2016), which attempts to rethink the design of blockchains, dividing them into parts network, consensus, storage, view, and side planes.

This general idea of ​​having identifiable blockchain subcomponents is taken and enriched by introducing the concept of the Gini coefficient. The Gini coefficient is the most commonly used measure of inequality, but it is based on several conditions that greatly reduce the value of the Gini coefficient in this case.

Srinivasan (2017) then generalized the minimum Nakamoto coefficient that measures the Gini coefficient of the subsystem, resulting in a final equal aggregate score for token holders of selected digital assets. This concept was widely welcomed and implemented in the public Python library, and it was further developed in the smallest Nakamoto coefficient "2.0".

secondary title

EOS Governance Status

2.1 Overview

Reiterates a core tenet of EOS governance: EOS token holders elect 21 Block Producers (BPs) from a broader pool of Block Producer candidates. These BPs follow and sign the "Peer-to-Peer End User License Agreement". But in reality, EOS governance has changed dramatically.

The first (provisional) EOS constitution was published in May 2018 and is administered by the EOS Core Arbitration Forum (ECAF), which also helps resolve disputes between EOS token holders. While ECAF and its “similar” approach to requiring BPs to sign and abide by the endorsement has drawn some criticism, its purpose is to balance the influence of BPs. This is seen as necessary, for example, when Reddit stated in 2019 that there is circumstantial evidence that EOS users believe block producers are too powerful. As one user put it, "Block Producers control all decisions made on EOS, from validating blocks to taking funds under your private keys."

Nonetheless, BP EOS New York galvanized previous efforts by publicly announcing that it would no longer follow ECAF's approach. Two dubious incidents ensued, the first being the first ECAF ruling to freeze 27 accounts without providing any reason, and the second being a popular but bogus ECAF ruling calling for the revocation of EOS transactions, but in clear contrast to previously agreed codes of conduct different.

The next evolution in EOS governance is not just ditching ECAF, but replacing the bylaws with the EOS User Agreement (EUA). Effectively, this means that at this point, all of the original guidelines for EOS governance are superseded. The process of replacing the charter is very confusing. The original proposal to replace the interim bylaws with an EUA called for a 15 percent voter participation rate. However, the final voter participation rate was only 2%. Despite this, EOS New York recommended to proceed and received the support of 15 BPs out of 21. Since then, 21 selected BPs and EUAs have become the core of EOS governance. There are some additional governance tools recently, such as EOS enhancement proposals and BP system upgrade proposals, but they only have support functions.

With a better understanding of the current and previous fundamentals of EOS governance, it is possible to move towards practical implementation. The way this is done is to assess people's biggest fears about EOS being "too centralized" in its governance. This is achieved by testing EOS against the three main goals of decentralization: collusion resistance, fault tolerance, and attack resistance.

Collusion resistance describes the ease with which system participants organize at the expense of others. It is arguably the most relevant metric for evaluating EOS governance.

A paper by Whiteblock (2019) found that block producer incentives for EOS were not properly structured, allowing the environment to facilitate collusion events. This is caused by:

• Block producer rewards come from inflation: currently, the entire inflation only involves BPs.

• 1 Vote 30 Votes: It facilitates transactions between votes. The community has discussed the 1 vote 1 vote (1T1V) solution, will it be implemented in the future?

• Vulnerable to Sybil attack: A single operation can allow multiple BPs to get more rewards. Can this be solved by 1T1V?

• Proxy: BP can control the proxy.

• Low turnout: This increases the weight of big players in voting dramatically.

• Tokens on exchanges: Hosted exchanges can use delegated tokens for voting.

In this situation, an economically rational agent must cooperate with others to preserve and maximize their profits. This situation is exacerbated by open, unregulated voting markets. Generally speaking, the process of buying votes is not a bad thing in itself, as any purchase can only be made if the buyer values ​​the vote more than the seller and the buyer is more interested in expressing a relevant position.

However, this could lead to an aristocracy, as the rich are in a "virtuous cycle" where they can accumulate more and more votes as they can use them to generate income. Linked to this, these 21 nodes earn almost 7 times more than other high-ranked BP candidates. In Radical Markets, Psoner and Welyl (2018) further evaluate the issue of buying votes and suggest that votes should obey a cost function, i.e. votes can indeed be traded and bought, but using or buying them reduces utility.

In EOS, however, there is no such cost function. In addition, the voting market is very opaque and structured through a large number of informal "quid pro quo" by BP. Still, it is possible to make a basic assessment of voting patterns.

2.2 Voting method

image description

Data source: EOSAuthority, Binance Research

image description

Data source: CoinMarketCap, Binance Research

Among the 164 giant whales, 84% (123) voted for 30 BPs at the same time, which is much higher than 52% (38,858) of the 30 BP accounts. The explanation for this situation is that large EOS holders may simply be more attractive targets in voting trading schemes. A second notable voting pattern is voting for ~20 block producers. The votes of ~20 BPs may also come from collusion and voting transactions. The revelation of the last point is that many (possibly a minority) voted for only one BP.

image description

Data source: CoinMarketCap, Binance Research

For example, the graph above shows that BP EOSHuobiPool has the most votes with 336 million votes. In comparison, the largest proxy “colintcrypto” controls about 10 million EOS, and the largest 21 proxies control 94 million votes. With the exception of BPs, the general idea behind proxies is similar to representative democracy, where voters choose selected individuals to act on their behalf.

Even this ignores that at least three of these proxies, such as bitfinexvp13, bitfinexvp21, and bitfinexvp33, are controlled by Bitfinex. This demonstrates the strong influence of Bitfinex. This is especially true when considering that eosrapidprod only needs 13 million EOS (approximately $70 million) to become the largest BP. But this analysis is too simplistic.

2.3 Fault Tolerance

A theoretical definition of fault tolerance can describe it as the number of failures a system can sustain while maintaining its functionality. There are many independent components that high redundancy generally increases fault tolerance. Depending on the chosen method of observing results, the fault tolerance of EOS can be evaluated by looking for indicators or events that describe faults and corresponding outcomes.

There are two prominent examples of BP failure:

• The first example is a BP failing to update the blacklist. The result of this failure was a loss of $7.2 million. Even if those funds were later recovered by Huobi, the EOS blockchain is not fault-tolerant.

• The second example is related to "misassignment" which forces multiple BP's nodes offline. These BPs were only temporarily replaced after 30 minutes, reducing the number of BPs and exposing the EOS blockchain.

In addition to these two high-profile events, BP can be evaluated by two other indicators: reliability and responsiveness.

Table 4 shows the reliability of the 21 block producers (as of February 10, 2020) by block and round availability. These metrics describe the number of blocks produced as the number of completed rounds divided by the number of planned blocks/rounds.

image description

Data source: AlohaEOS, Binance Research

image description

Data source: AlohaEOS, Binance Research

Except for Bitfinex, EOS WIKI, and EOSHuobiPool, all 21 BPs had short and consistent execution times, indicating that their BPs' activities were allocated by sufficient resources. However, what is interesting is that the two exchanges that performed worst among BPs performed well in terms of reliability and responsiveness.

2.4 Anti-attack

The last evaluation goal of EOS governance is the ability of EOS to resist attacks. This attack resistance can be manifested in various ways such as censorship attacks. In general, decentralized systems are said to be more expensive to attack since there is no centralized point of failure.

However, circumstantial evidence suggests that the EOS network has been significantly consolidated. For example, EOS New York reported that one entity registered 6 different BPs. Likewise, many BPs vote for themselves through proxies. Some of these proxies are publicly associated with a particular BP - for example, Huobi has 5 proxies with a total of 5 million EOS, BigOne has 15 proxies with a total of 1 million EOS, Bitfinex has 13 proxies with a total of 4000 Ten thousand EOS. Additionally, BPs may operate as many as 50 different covert agents.

Another attack vector is geographical in nature. One-third of block producers are in China, and more than half are in Asia. So there is a strong regional focus, similar to the geographic distribution of mining pools. Nonetheless, EOSAuthority's correlational analysis reveals that there are two main clusters. One around EOSAuthority and the other around EOSHuobiPools. This reflects the geographical differences mentioned earlier.

secondary title

no set questions

EOS has a feature called "regproducer" which is a "mutually agreed guideline" for enforcing on-chain standards among block producers. A BP submits a referendum to update the contract, thereby raising the BP threshold. Since any change must be approved by at least 15/21 BPs to be implemented, and this update received only 13 out of 21 votes, it was doomed not to be implemented.

However, large proxies may change the ranking of candidate block producers by redistributing their votes. This is what happened in the case described earlier, causing the two candidate block producers to rise in the ranking, become BPs, and thus gain voting rights.

Therefore, it can be concluded that BP may have the sole decision-making power, but still rely on voting to decide whether to become and retain the right of BP.

Therefore, large proxies and accounts have considerable influence. Similarly, Block.one launched EOS, has over 96 million EOS, and can theoretically use it to vote and change the order of BPs at any time. In terms of quantity: the number of EOS held by Block.one is almost 10 times that of the agents who affect the ranking of the 7 BPs and BP candidate companies. Likewise, Block.one and the next four largest EOS holders already account for a quarter of the entire circulating supply.

Another promising approach to improving EOS governance could involve formalizing the process of determining voting transactions. By introducing a formal, transparent mechanism with a cost function for voting, informal coordination can be reduced by introducing diminishing returns for buying EOS votes. This would preserve the ability of large token holders to have greater influence, but avoid further EOS' "virtuous circle".

Other previous proposals to improve EOS include:

• Includes random shuffling to select 21 BPs from the 100 largest BP candidates.

• Introduces general inflation based on the amount of EOS staked.

• Introduce downvotes and voting caps.

• Proxies with Block.one funds are introduced to vote based on the community's preference for BPs.

• A BP diversity criterion is introduced to limit regional concentration.

in conclusion

in conclusion

"While Web2 is defined as "move fast and break things," Web3 should follow ideas like "do things the right way."

—— Andrew Keys, DARMA capitalization

From a practical point of view, Keys' inspiring call can be understood as a basic encryption economic need: to complete it as early as possible before the network is established.

Having said that, it remains to be seen whether EOS can overcome its own structural problems. Unfortunately, the vulnerability of EOS to Sybil attacks reduces transparency, making it difficult to draw definitive conclusions about the voting patterns of BPs and BP-related proxies.

However, there are two unrelated issues that may indirectly exacerbate the situation for EOS governance, and they relate to dApps on EOS. While EOS provides a variety of documentation for developers, only a few APIs are available. The provision of expensive APIs is entirely voluntary and does not create any obligation for BP. At the same time, the number of users of EOS dApps has dropped significantly in the past 6 months. That being said, the upcoming beta launch of "Block.one's Facebook", Voice, is widely considered a milestone.

While the extent of collusion between commodities is often unclear, circumstantial evidence points to problems with network integration. This seems to stem from the rationale behind EOS: governance is opaque and little is known about the voting market, which exacerbates the incomplete distribution of incentives using the dPoS consensus system.

However, this problem is not limited to EOS, it seems to be an inherent problem with dPoS blockchains, and is often exacerbated by the escrow ownership of tokens. As the largest dPoS blockchain, EOS naturally encounters a high level of scrutiny, so groundbreaking solutions must be identified and adopted.