A Retrospective of the EOS Token Sale
A detailed look at some noteworthy aspects of the EOS token sale and an analysis of its mechanics alongside other sales during the same time period.
Danning Sui, Johannes Pfeffer — Alethio Data Science
Jeff Gillis, Everett Muzzy — ConsenSys
Alethio’s block explorer and analyzer is live. Explore the Ethereum network at EthStats.io.
Background on Alethio
At Alethio, which is a ConsenSys formation, our mission is to make blockchain data accessible for a variety of use cases. Through data, we can empower users of blockchain systems to further understand transactions and behavior, leading to increased recognition of how these systems work — a critical tool for both retrospective thinking and prospective planning by the various actors in the systems.
Our proprietary analytics platform, fueled by data science and semantic modeling, is useful in achieving real time data analysis but it is equally useful in retrospective study. We call this type of retrospective analysis blockchain archeology.
Scope
Over the last two years or so, we have seen Token Sales evolve into an important element of the crypto and blockchain ecosystem. The issue of price volatility aside, they have allowed legitimate projects and teams to raise money, build a network effect, and gain market recognition. We’ve also seen exaggeration and fraud. Our nascent industry is still in the process of establishing standards and best practices for token sales. With Alethio’s data analytics platform, we can tap into the transparency of the Ethereum network to investigate high-profile token sales and help the community understand their mechanics. By helping entrepreneurs, investors, regulators, and enthusiasts understand the nuances behind token sales, we aim to contribute to a more equitable, transparent, and secure decentralized future.
Why EOS?
EOS.IO [1] is a blockchain platform created by the company block.one. To fund this new blockchain, block.one launched a year-long, uncapped token sale on the Ethereum blockchain by issuing an ERC20 token known as the EOS token. The EOS token sale is the largest crowdfunding event ever as of the time of this writing [2]. By the time the token sale ended in June of this year, EOS received over $4,000,000,000 in contributions and had risen to one of the highest market caps of all blockchain systems at over $10 billion implied value [3]. Moreover, in October 2017, block.one promised a third-party audit of its token launch[4][5], for which the blockchain community is still waiting [6][7][8]. Today, all EOS tokens are locked on Ethereum and have been “transferred” to the EOS blockchain.
Given the publicity, market cap, and promised audit of the EOS token sale, we believe it to be a worthy subject for some archaeology. We hope that through these expeditions, we can unearth data that allows us to investigate and suggest best practices of token sales in general. Below, we will present some unique characteristics found from the EOS token sale with original data and visualizations.
Note: in this post, we will use token sale, token launch, and crowdsale interchangeably.
Methodology
Because the EOS token launch lasted one year (June 26, 2017 to June 1, 2018), we collected and analyzed data between block 3,904,416 and block 5,810,294. In these roughly two million blocks, we examined the flow of funds (Ether and EOS tokens) to and from the EOS Crowdsale Contract and studied their origin and destination addresses.
Specifically we have looked at the following aspects:
- All messages sent to the Crowdsale Contract (messages are transactions plus resulting smart contract actions).
- All EOS token transfers resulting from these messages.
- The transfer timeline of funds between the EOS Crowdsale Contract and the EOS Funding Wallet.
- Addresses that received ETH from the Funding Wallet, and subsequent flow of funds.
There are some areas where we do not possess all the information needed to make definite conclusions on some of our arguments throughout this piece, which is an academic exercise. We make it clear when assumptions are made, and we apply our best objective judgement to areas with unclear or incomplete data. We encourage response and collaboration from the blockchain community, including and especially the EOS community. Readers will note that many data gaps occur when information moves off-chain or to centralized exchanges without transparency, while data abundance occurs where transactions are decentralized and conducted on-chain.
Unique Aspects of the EOS Sale
The EOS crowdsale displayed some uncommon attributes. We especially looked into the effects of the following features of the sale:
Fund Withdrawals Throughout the Sale
During the EOS token sale, funds were permitted to be transferred out of the token sale wallet before the token sale ended. Just five days after the beginning of the token sale, block.one first withdrew funds from the token sale wallet, and continued to withdraw throughout the crowdsale. We note that this is an unusual feature of the token sales we’ve studied, with most prohibiting transfers out of the receipt address until completion of the sale either through code or contract.
Token Sale Length
The EOS sale ran for a period of one year. A one year sale is significantly longer than other well-regarded token launches during the same time frame that exhibited significantly shorter windows for participation.
Token Trading Availability on Exchanges
Due to the length of the sale and the fact that EOS tokens were distributed to purchasers while the sale still commenced, EOS tokens were available for trading on centralized exchanges for the duration of the token sale [9]. The availability of tokens of an ongoing crowdsale on centralized exchanges is uncommon, according to our research.
Centralized Token Holding
EOS’ Constitution states that no user shall own more than 10% of token supply [10]. Adherence to the Constitution, however, is impossible to verify because multiple wallets could be owned by the same party, which would not be apparent based simply on blockchain addresses.
With original data, visualizations, and insights, let’s dive into the EOS and other token sales to learn more about the mechanics and strive towards the derivation of some token sale best practices.
Part 1: Token Sale Length and Withdrawals
We analyze these features in tandem, as one appears to facilitate the other.
The Virtue Poker Token Sale: A Representative Baseline
In order to compare the EOS token sale against others, we analyzed the Virtue Poker Token Sale (disclosure: VirtuePoker is a ConsenSys formation), which lasted from April 25 to May 9, 2018 (shown in Figure 1). The green block labeled VPP indicates the length of Virtue Poker’s crowdsale. The first vertical red line to the right of VPP marks the first transfer of funds out of the sale wallet. Each vertical red line afterwards marks another withdrawal. The lines only indicate the point in time of a withdrawal, not the magnitude. The VPP sale ended on May 9, but funds were not first withdrawn until May 30. This type of visual will also be used for the other token sales below.
Comparing EOS’ Withdrawal Practices to Other Token Sales
In general, to participate in a token sale, contributors transfer funds to a crowdsale contract. In the case of EOS, contributors deposited their funds to a crowdsale contract address that we label as the EOSSale contract.
A crowdsale contract usually transfers funds to a designated “funding wallet” owned by the company or project which is supposedly more secure and/or has more wallet-like features. In the case of EOS, funds were transferred from EOSSale to a funding wallet labeled as EOS-Owner.
Typically, the funds stay in this funding wallet until after the crowdsale ends (as shown by the vertical red lines in figure 1). Projects will often spend this time — between the end of the token sale and the first withdrawal of funds — performing crucial tasks such as setting up the company to receive the funds, ensuring valid know-your-customer (KYC) on all contributors, or working with exchanges to make the tokens available for trading.
After the crowdsale ends, the company typically transfers the contributed funds out from the designated funding wallet to other accounts. In this article, we refer to the movement of funds from the funding wallet as the owner, company, or project withdrawing funds.
EOS’ Withdrawal Parameters
The EOS token sale between June 26, 2017 and June 1, 2018 followed a structured, predetermined release schedule according to the purchase agreement [11]. In total, 1 billion EOS tokens were distributed, with 10% of them reserved for Block.one. For the first 5 days of the crowdsale (June 26 — July 1, 2017), the contract distributed 20% of the tokens (200,000,000). The remaining 70% were split evenly into 350 portions (2,000,000 tokens each), and distributed in 350 consecutive 23 hour periods (July 1, 2017 — June 1, 2018).
Figure 2 shows token launches that overlapped with the EOS sale and lasted at least two weeks. The EOS sale is shown at the bottom. The red lines indicate the points in time when funds were withdrawn from the token sale funding wallet. The original data for this illustration is attached in the Appendix.
While all other observed projects waited until after the end of their sale to first withdraw funds, block.one/EOS transferred funds out of the funding wallet continuously throughout the sale. Block.one first withdrew funds on July 1, 2017, just 5 days after the EOS token sale began, which is also the day that EOS started trading publicly on Bitfinex [9].
In total, EOS withdrew funds 99 times, 93 of which occurred within the token sale time frame. Those 93 transfers accounted for approximately 7.75 million ETH withdrawn from their funding wallet, which is 90.8% of the total funds raised through the sale. In this time window, funds were withdrawn, on average, every 3.8 days.
There are no written rules or laws regulating withdrawal of funds during a token launch. There are, however, obvious concerns about how a token sale owner could use the withdrawn funds [12]. For example, it is conceivable that funds could be transferred out of the token sale wallet during the sale and then used to buy the token again, perhaps after being routed through a few steps such as a centralized exchange. This could result in artificially inflated demand for EOS by raising its market price and fuelling speculation and interest in the sale. We look at this possibility in more detail later in the article.
Given this possible concern, we took a closer look at where the funds went during and shortly after the crowdsale.
Part 2: Flow of Funds and EOS
Who Owns EOS Tokens?
The Ethereum-based EOS tokens were frozen on June 2nd 2018. We looked at the top ten addresses with the most EOS tokens at time of freezing (table 1). Collectively, these ten addresses account for almost 50% of total EOS supply; but as discussed below, five out of the ten appear to be exchange addresses.
The top address was created by the EOS contract and is non-transferrable [13]. This 10% is reserved for Block.one and cannot be reduced or increased according to EOS Token Purchase Agreement [11].
The address with the third most EOS tokens had a high transaction volume with Bitfinex_Wallet4, the main active wallet of the Bitfinex Exchange (which also happens to be one of the 21 block producers) [14]. As shown in Table 2 below, Bitfinex_Wallet4 transacted over 1 million ETH to the third highest EOS wallet during the token sale time frame.
When we first noticed the high transaction volume between the #3 addresses and the Bitfinex_Wallet4 address, we only made the conclusion the two addresses had a close “relationship”. Since our original analysis, however, the #3 address has been identified on Etherscan as Bitfinex_Wallet5, Bitfinex Exchange’s newest main wallet. There is no absolute way to prove the classification of the wallet, but given its label on Etherscan, we continued our analysis under the assumption that address 0x742d35cc6634c0532925a3b844bc454e4438f44e, the #3 address, is Bitfinex_Wallet5, which is controlled by Bitfinex Exchange. To restate the situation given this assumption, therefore: during the course of the EOS token sale, Bitfinex_Wallet4 transferred approximately 1 million ETH to Bitfinex_Wallet5, which has the third most EOS tokens of any address. Both addresses are controlled by Bitfinex Exchange.
Similarly, addresses #5 (0x6cc5) and #6 (0xfe9e) have also recently been labeled on Etherscan as OKEx’s and Binance’s main wallets, respectively. We will continue our analysis under the assumption these labels are correct. To clarify, this does not mean the exchanges contributed directly to the token sale, but only demonstrates that as of August 9, 2018, they were in the top 10 EOS-owning addresses. It is important to understand that the ownership of these tokens are likely split between the exchanges themselves and their customers. Without further information from the exchanges, one cannot know which part the exchanges own and which part the customers of the exchange own.
When we investigated wallet #7, we identified it as the cold wallet of PXN, which is an exchange platform [15] that advertises a close business relationship with Yunbi and BigONE, two exchanges founded by a single individual who also invested early in Block.one.[16] Furthermore, address #8 exhibits an active token transfer relationship with the BigONE hot wallet, and we will operate under the assumption addresses #7 and #8 may belong to the BigONE exchange. Thus, the exchanges PXN, BigONE, and Yunbi — and the customers’ EOS tokens on those exchanges — may control altogether 5.77% (added from #7 and #8), which is more than it looks like on the surface (3.24% and 2.53%).
The remaining unlabelled addresses (#2, #4, #9, #10) account for an accumulated value of over 170M EOS, roughly 17.5% of all EOS tokens. All of these addresses are external accounts, meaning they have no associated smart contract code. Due to the pseudonymity of external account addresses, their identities are opaque. The EOS community may have more insight into the identities of these remaining top addresses, or the EOS audit may provide more information.
Following The Bitfinex Trail
We were curious to find out where the ETH from the crowdsale flowed after it had been transferred out of the funding wallet. To visualize the flow of funds, we created a Sankey diagram based on the direction and value of the transactions. The green block on the far left side is the EOSSale Contract, which received 7.21 million ETH from contributors. From the crowdsale contract, all of the funds were transferred into the “EOS-Owner contract” — the red section to the right. This contract was the funding wallet, from which ETH was transferred to other wallets.
The diagram shows the payouts from the EOS funding wallet beginning five days after the token sale began. In the first “layer” of transfers (red), funds are withdrawn to a total of 21 recipients in varying amounts. On Figure 4, these 21 addresses are the first red “blocks” that funds hit once they begin moving from EOS-Owner contract on the far left area in red on the graph.
We were able to follow the direction of transactions after these initial 21 recipients and continued our analysis into deeper layers of the EOS-originated funds. Each block represents an origin/destination address, and a larger block or stream means a greater amount of funds being transferred. As we followed further, we identified the main wallets of more and more exchanges being involved in funds transfers. This is not atypical of expected behavior; as more addresses transact on the network, they do so increasingly through exchanges, which serve as network activity hubs.
We did find significant activity, however, between the EOS-Owner address and the Bitfinex4 address — shown as the yellow block in the diagram. Over four million ETH have flowed from the EOS funding wallet to multiple individual wallets, and then to Bitfinex’s main wallet (Bitfinex4) alone. We separated the transactions flowing from the EOS funding wallet to Bitfinex4 and compared the timeline and volumes with market fluctuations in the price EOS.
Payouts to Bitfinex occurred along the following timeline between December 2017 and June 2018:
Daily transaction volumes peak in May and June, which raises the stock of Ether to 2 million then 3 million, respectively, as shown by the horizontal lines. This was a huge amount of ETH and possibly related to the sale of 180,000 ETH in April and May on Bitfinex [17].
Circular Flow: The Case for Smart Contracts that Promote Trust
In this section, we’re applying the objective data we’ve analyzed to a subjective opinion of governance. There are many ways one can design a crowd sale and the different options are still being tried out constantly. As they are tried, their respective advantages, shortcomings, and lessons emerge. As we analyzed the EOS sale, we identified some problematic patterns that we’d like to discuss as potential sources of manipulation of exploitation. Again, we’d like to stress that we didn’t see any conclusive evidence that this actually happened in the EOS sale.
Two properties of a token sale that are intrinsically problematic in our opinion, are the following:
- The token sale allows the funds raised to be taken out before the end of the token offering.
- The token sale allows tokens to be moved (i.e. resold and traded in the secondary market) before the end of the sale.
Both of these properties create an incentive to the conductor of the sale to engage in practices that promote price increase and speculation. Specifically, in token sales with these properties, there exist incentives to engage in the following:
The token sale allows the funds received to be taken out before the end of the token offering
Incentivised action: Take a fraction or all of the funds raised and use them as input into the sale. Since the issued tokens go directly back to the sale holder, this can be repeated many times with only minor transaction costs.
Incentive mechanism: Such action is incentivized because it simulates high activity in the token sale and allows for the manipulation of the price by the issuer. If a higher percentage of the issued tokens are owned by the entity that controls the sale, they have more influence over the supply of tokens on the market and thus the trading price. This influence would be hard to perceive because the entity may appear as smaller, independent token holders. In addition to price increases, an increase in transaction volume may also drive speculation, even if artificially created.
Observability: This type of funnelling can only be detected if funds are moved directly, or over a few hops to the sale contracts. It can be easily concealed by routing funds through opaque hubs, such as an exchange or a dedicated tumbler before funneling them back into the sale.
The token sale allows tokens to be moved (i.e. resold and traded in the secondary market) before the end of the sale.
Incentivized action: Take a fraction or all of the funds raised and use them to buy the token sold in the crowd sale on the market, while the sale is still going on.
Incentive mechanism: This creates an artificial demand for the token, and with a delayed supply, the issuer can pump up the price and stimulate greater demand during the token sale. The evolving token price is meant to be an indicator for the long run potential of the project, or the long run transaction demand for the currency. It is not meant to be manipulated by whales with deep pockets, especially those that benefit from the price impact their purchases have on raising the price, such as token issuers.
Observability: This price pumping scheme is very hard to detect because most token trading occurs on opaque centralized exchanges. A possible indicator can be large price movements after raised funds are moved to exchanges — but this can be obscured. Usage of decentralized exchanges that exist entirely on-chain, such as AirSwap (disclaimer: AirSwap is also a ConsenSys formation), can eliminate this black-box by forcing all activity to the transparent ledger.
Conclusions
Our analysis of the EOS token sale observed token mechanisms, transaction activity, and exchange funding on a scale that is relevant for the Ethereum ecosystem. There is no law, regulation, or codified best practice classifying early redistribution of funds during a token sale as illegal or malicious. However, it is not a common practice compared to other token sales we have studied or heard about, and it can incentivize inflation of the crowdsale and the token price. Investors/contributors to a token sale should understand the smart contracts under which the toke sale is engineered. More importantly, token sale teams should make that information readily available. If there is no smart contract stipulation about how the contributed funds are treated during the token sale, contributors can assume this behavior could occur.
It is easy to avoid the poorly incentivized mechanics described above by designing a crowdsale so that funds cannot be withdrawn and tokens are not paid out before the end of the sale. Smart contracts provide the means to do transparent sales where a large portion of the required trust is externalized into the contract. This power should be used.
Further measures to increase the trust that no funneling schemes are intended by the sale operators would be locking the raised funds in time vaults that allow only fixed withdrawals per time unit (e.g. 2% of raised funds per month). Additionally, the tokens allocated to the sale operator can be time locked for a certain amount of time (e.g. first 6 months after the sale). These are well known mechanisms, similar to some corporate vesting schemes, and have been used in past token sales. Not employing them undermines trust and is neglectful to the investors, in our opinion.
As part of block.one’s promised third-party audit of the token sale, a call to share information is justifiable, elicited by the Ethereum, EOS and blockchain communities. This call requests information regarding what was done with the funds withdrawn early and often during the token sale in order to confirm whether block.one was trading the token itself or making markets for EOS during the sale.
We readily admit that our research was incomplete in areas, and we used and clearly labeled assumptions we made. As a data analysis resource, Alethio has a strong commitment to staying neutral by providing facts and best judgements based off objective and/or verified information. By creating a quantified context we hope to encourage transparency and a more educated, collaborative community.
Special Thanks to Joseph Lubin, Matt Corva, Amanda Gutterman, Dominik Muhs.
References
[1] EOS.IO — Wikipedia https://en.wikipedia.org/wiki/EOS.IO
[2] A blockchain start-up just raised $4 billion without a live product https://www.cnbc.com/2018/05/31/a-blockchain-start-up-just-raised-4-billion-without-a-live-product.html
[3] Rank of Coin According to Coinmarketcap https://coinmarketcap.com/coins/
[4] EOS FAQ https://eos.io/faq
[5] EOS promised an audit in October https://www.reddit.com/r/CryptoCurrency/comments/8oj7ct/eos_promised_an_audit_in_october_to_prove_that/
[6] Where is the audit? https://www.reddit.com/r/eos/comments/8xnsd9/where_is_the_audit/
[7] What happened to the ICO audits? https://www.reddit.com/r/eos/comments/8gkbf8/day_2_what_happened_to_the_ico_audits/
[8] Still No Audits https://www.reddit.com/r/eos/comments/7uni7v/still_no_audit/
[9]EOS is available on Bitfinex since July 1, 2017 https://steemit.com/eos/@cob/making-sense-of-the-eos-ico-in-laymen-s-terms
[10] EOS Constitution — Article V — No fiduciary htps://github.com/EOSIO/eos/blob/37ce45c0b60d2710569c2d1a9229945cc0e855a9/governance/constitution.md
[11] EOS Token Purchase Agreement — Exhibit A section https://eos.io/documents/block.one%20-%20EOS%20Token%20Purchase%20Agreement%20-%20September%204,%202017.pdf
[12] Anyone noticing the EOS red flags? https://www.reddit.com/r/ethtrader/comments/7dbev3/anyone_noticing_the_eos_red_flags/dpx8m2x/
[13] EOSSale Contract Code — search “0x1b” https://etherscan.io/address/0xd0a6e6c54dbc68db5db3a091b171a77407ff7ccf#code
[14] EOS Block Producer https://medium.com/@bensig/eos-block-producer-faq-8ba0299c2896
[15] PXN Exchange Platform https://www.pxn.one/
[16] Investment Profile of Xiaolai Li https://pseps.com/person/xiaolai-li
[17] EOS Has Spent One Billion Dollars Worth of Eth in the Past Month Data Reveals https://www.trustnodes.com/2018/05/09/eos-spent-one-billion-dollars-worth-eth-past-month-data-reveals
Appendix — Token Sale Detailed Information
(data pulled on Aug 9, 2018)
Alethio’s EthStats blockchain explorer and analyzer is live!
Disclaimer: The views expressed by the authors and contributors above do not necessarily represent the views of Consensys AG. ConsenSys is a decentralized community with ConsenSys Media being a platform for members to freely express their diverse ideas and perspectives. To learn more about ConsenSys and Ethereum, please visit our website.
