Fighting Vampires: The Summer of DeFi and the Birth of the UNI Governance Token

Written by: Sizheng Fan, Tian Min, Xiao Wu, Cai Wei

Compiled by: tiao

Introduction

Recently, Erin Koen, the governance lead of Uniswap, initiated a proposal in the forum to improve the governance system and fee mechanism of the protocol. He proposed that users who stake and delegate UNI governance tokens should receive a proportional share of the protocol fees to incentivize and strengthen Uniswap's governance. This proposal was well-received by the UNI community, and the price of the UNI token surged by 60% in a single day.

The author noticed that within the LXDAO community, as well as in the broader web3 community, another point of concern emerged regarding this proposal: it also involves an adjustment of fee distribution and incentive mechanisms, where holders of UNI tokens will benefit from this change.

For this LXDAO translation, we chose to publish the research article "Towards understanding governance tokens in liquidity mining: a case study of decentralized exchanges," authored by Sizheng Fan, Tian Min, Xiao Wu, and Cai Wei from The Chinese University of Hong Kong, which was published in July 2022. Due to time and capacity constraints, the author has omitted the "Method" section, most of the data charts, and references from the original text, and made some edits to the titles for readability, while striving for accurate translation and adding reminders where relevant.

In the study, the authors used data analysis to recreate the "vampire attack" of the summer of DeFi in 2020, which was precisely in response to the attack from SushiSwap, leading to the birth of the UNI governance token. We can see that governance tokens have had two purposes since their inception—governance and incentives—entangled between the ideals of web3, commercial competition, and incentive mechanisms. It is believed that this article will help everyone understand the background and motivation of Uniswap's proposal and prompt further thinking about broader issues and possibilities in web3.

1. Article Summary

Liquidity mining is a hot topic in the decentralized finance (DeFi) community, significantly boosting the total value locked (TVL). To attract users, most decentralized applications (DApps) distribute governance tokens in liquidity mining.

However, the effectiveness of this approach still lacks in-depth evidence. This article explores governance tokens in liquidity mining through a typical case: SushiSwap attracted a large amount of Uniswap's liquidity in the short term by copying Uniswap's code and issuing governance tokens in advance.

We collected over a year of trading data from Uniswap and SushiSwap and conducted a detailed analysis of the behavior of liquidity providers (LPs). We also designed a scalable unsupervised clustering method to construct similarity graphs based on trading flow metrics, revealing patterns of liquidity providers with similar behaviors. These patterns ranged from inactive and cautious liquidity providers to those providing minimal liquidity and those pursuing risk in the short term.

Based on this, we discussed the impact of governance tokens on liquidity mining and explained their appeal to users through their influence on behavior and decision-making.

2. What is Liquidity Mining?

Liquidity mining is a mechanism that incentivizes users to provide liquidity for decentralized exchanges (DEXs), first proposed by IDEX in 2017 and further refined by Synthetix and Chainlink in 2019. Mid-2020 saw an explosive growth in liquidity mining, with projects like Compound, SushiSwap, and Uniswap joining in. Liquidity mining has become an essential part of the DeFi community. According to DeFi Pulse data, the total value locked (TVL) in the market was $1.05 billion in early June 2020. By September, community efforts had increased the TVL tenfold. This astonishing growth and enthusiasm rivaled the boom of initial coin offerings (ICOs) in 2017.

Liquidity mining is a mechanism that incentivizes users to provide liquidity for DEXs, similar to bank deposits. Users can deposit token pairs in AMM-based DEXs, enabling trading without an order book through smart contracts. Traders can swap one token for another on the DEX. After each trade, the DEX charges a certain trading fee, which is distributed as interest rewards to liquidity providers. Additionally, liquidity providers can earn governance tokens as a stake in the protocol to encourage their participation. Governance tokens are a type of cryptocurrency that grants voting rights representing blockchain projects. In recent years, governance tokens have been widely used in DeFi projects to facilitate decentralized decision-making by users. Governance tokens also have economic value and can be traded on centralized exchanges (CEXs) and DEXs. The issuance of governance tokens is regulated by the protocol and typically follows a decreasing trend.

Currently, most decentralized protocols have adopted liquidity mining as an innovative and efficient decentralized approach. By increasing governance token rewards in liquidity mining, protocols can attract more liquidity providers, as seen in SushiSwap's "vampire attack" on Uniswap. However, the effectiveness of this approach still lacks in-depth analysis.

3. Using Blockchain as a Method

To answer how governance tokens affect users' strategies, a direct method would be to conduct user surveys. However, in the blockchain field, this method is costly and difficult to implement. Obtaining representative survey results requires collecting a large amount of user information. Sending questionnaires through wallet addresses or filtering users participating in governance tokens from the blockchain community is not easy. Moreover, using cryptocurrency as a reward can impose a significant economic burden. Therefore, we chose a more prudent approach by sourcing the information we needed from open-source data.

This article explores the impact of governance tokens on liquidity mining through a specific case: SushiSwap attracted a large amount of liquidity by issuing the governance token SUSHI and permanently incorporating it into liquidity mining rewards, thereby launching a "vampire attack" on Uniswap. Subsequently, Uniswap also issued the governance token UNI and used it as a liquidity mining reward from September 18 to November 18, 2020. We analyzed the impact of governance tokens on liquidity mining based on the differences in the issuance time and reward cycles of governance tokens between Uniswap and SushiSwap in this case.

1. Macroscopic Level: We analyzed the macro data of Uniswap and SushiSwap, exploring the impact of external factors such as ETH prices and incentive policies on total value locked (TVL), function calls, and user numbers.

2. Microscopic Level: We conducted unsupervised clustering of liquidity providers and compared behavioral changes before and after the vampire attack to reveal the sensitivity of different types of liquidity providers to incentives.

4. Findings

Our main contributions are as follows:

Database Construction: We briefly introduced significant events of the two decentralized exchanges, Uniswap and SushiSwap. Uniswap is the most popular DEX on Ethereum, while SushiSwap is one of the earliest and most prominent forks of Uniswap. Although the transparency of blockchain ensures the open-source nature of on-chain data, integrating, preprocessing, and extracting usable data remains a daunting task. We collected nearly a year of granular trading data covering records from approximately 300,000 addresses across these two exchanges to construct the database.

Trading Flow Extraction and Clustering: We proposed a method for formatting address trading flows and implemented it based on our database. We also utilized an unsupervised hierarchical clustering method to classify liquidity providers into six categories based on their behavioral characteristics: optional, lightly active, lightly inactive, risk-averse medium, risk-seeking medium, and risk-seeking heavy.

Result Analysis: The clustering results show that governance token rewards can attract more liquidity providers in the short term, but subsequently, these liquidity providers often withdraw liquidity to pursue higher income, leading to a decline in annual percentage yield (APY) and causing losses, especially for medium and heavy liquidity providers. Such behavior contradicts the original intent of governance tokens and can trigger a vicious cycle, making it easier for traders to encounter slippage, further reducing liquidity providers' earnings. Additionally, by comparing the overlapping liquidity provider ratios of SushiSwap and Uniswap over the year, we found that under the long-term incentives of the SUSHI governance token, SushiSwap gradually cultivated a dedicated group of stable liquidity providers.

5. Background

5.1 Ethereum and Modern Cryptocurrencies

Ethereum is a blockchain platform based on innovations from Bitcoin, providing developers with an end-to-end system for building mainstream software applications. It pioneered a new computing paradigm: a trustful object messaging compute framework. Smart contracts are distributed scripts that execute synchronously across multiple nodes on the blockchain without requiring external trust institutions.

Cryptocurrencies based on Ethereum follow the ERC-20 standard, sharing several common characteristics, including tradability, which is a premise for this article. They have a fixed and limited token supply and utilize the Ethereum blockchain, where anyone can view them publicly. Additionally, token owners can freely transfer control of the tokens. These features have facilitated the formation of a market where users can trade tokens, whether through exchanges (including both decentralized and centralized exchanges) or peer-to-peer.

5.2 Impermanent Loss

As mentioned above, liquidity mining is simply a passive income method that helps cryptocurrency holders profit by utilizing existing assets rather than letting them sit idle in wallets. Assets are deposited into decentralized exchanges, and in return, the platform proportionally distributes the fees earned from trading to each liquidity provider. Impermanent loss, also known as divergence loss, refers to the losses incurred when funds are exposed in liquidity pools. This loss typically occurs when the token ratios in the liquidity pool change, meaning users suffer negative returns compared to simply holding their tokens without depositing them into the liquidity pool. In such cases, DeFi protocols tend to use trading fees paid by traders to compensate liquidity providers. Some even add extra rewards—governance tokens—to attract more liquidity.

5.3 Development History of Uniswap and SushiSwap

The development of Uniswap and SushiSwap can be divided into four stages: steady growth, vampire attack, Uniswap's counterattack, and the boom period.

The first version of Uniswap, Uniswap V1, went live on the Ethereum mainnet on November 2, 2018. The initial liquidity of this version was only $30,000, consisting of three tokens. However, the design of Uniswap V1 only supported automatic exchanges between ETH and ERC-20 tokens, meaning each liquidity pool had to include ETH. Therefore, trading between ERC-20 tokens required ETH as an intermediary, increasing gas fees, commissions, and slippage. In other words, swapping tokens required two transactions instead of one.

Steady Growth Period (May 19, 2020 - August 28, 2020): In May 2020, Uniswap released its second version, which primarily featured direct exchanges between ERC-20 tokens, significantly reducing trading costs and time, as well as the risks of impermanent loss faced by liquidity providers. Additionally, Uniswap V2 introduced new features such as on-chain price oracles and flash trading.

Vampire Attack (August 28, 2020 - September 17, 2020): At the end of August, SushiSwap entered the market as a clone of Uniswap but offered governance token rewards to Uniswap's liquidity providers, aiming to shift Uniswap's liquidity to its own platform and compete directly with Uniswap. Specifically, the first step of the vampire attack was to reward Uniswap's liquidity providers with SUSHI tokens, requiring them to stake their UNI-V2 tokens on SushiSwap. SushiSwap established an incentive plan for SUSHI tokens: 1,000 SUSHI would be distributed to Uniswap's liquidity providers for each Ethereum block, covering multiple liquidity pools. Once sufficient liquidity was transferred, the staked UNI-V2 tokens would migrate from Uniswap to SushiSwap. Ultimately, SushiSwap not only took away liquidity but also Uniswap's trading volume and users.

Uniswap's Counterattack (September 18, 2020 - November 18, 2020): To counter SushiSwap's vampire attack, Uniswap launched its own token, UNI, on September 16. Surprisingly, a portion of UNI was retroactively allocated. Addresses that interacted with Uniswap before September 1 could claim 400 UNI, worth approximately $1,200 at the time. Furthermore, Uniswap created four liquidity pools and incentivized liquidity providers with additional UNI tokens over the next two months, resulting in millions of dollars in increased liquidity.

Boom Period (November 18, 2020 - May 19, 2021): After ceasing the distribution of additional governance tokens, Uniswap's total value locked (TVL) experienced a brief decline. However, starting in November 2020, as the prices of Ethereum and Bitcoin rose, the trading volume and total value locked of Uniswap and SushiSwap rapidly increased as more funds flowed into the entire DeFi ecosystem.

6. Recreating the Vampire Attack

6.1 Dataset

To study the incentive effects of governance tokens in liquidity mining, we used the case of SushiSwap's vampire attack on Uniswap and Uniswap's counterattack, obtaining public records of addresses that interacted with the contracts of these two platforms from Etherscan between May 2020 and July 2021. Based on this data, we made the following discoveries:

Although Uniswap's trading volume is ten times that of SushiSwap, the proportion of liquidity provider addresses on SushiSwap is twice that of Uniswap. This may indicate that SushiSwap's perpetual governance token distribution can enhance liquidity provider participation.

While both Uniswap and SushiSwap offer the same services, the proportions of users utilizing these services differ. On Uniswap, users are more inclined to trade tokens through ETH, whereas on SushiSwap, users are more likely to engage in direct token-to-token trading.

6.2 Contracts and Functions

The router contracts of DEXs are integrated interfaces used to swap different token pairs or manage liquidity. Therefore, we can track the behavior of all DEX users through the router contracts.

To achieve these functions, the router contracts contain various functions related to token trading and liquidity, which are generally implemented as interfaces that call core contracts to perform specific operations. In our dataset, the Uniswap core contract was called 26 times, while the SushiSwap core contract was called 33 times, with 21 of them being the same.

Based on their functions, these functions can be roughly categorized into three types:

"Swap" functions, used to facilitate trading between ETH/token or token/token under different conditions; "Add/Remove" functions, used to increase or decrease liquidity in terms of ETH or token pairs; and "Utility" functions, used for querying, managing, or emergency responses.

To accommodate all possible trading scenarios, these functions generated different variants according to specific naming conventions, as shown in the figure below. Thus, we can filter addresses that used specific functions by filtering keywords.

Figure 1: Function Naming Rules

6.3 Data Collection

We retrieved external transaction records before July 2021 in block order. Then, we grouped external transactions by wallet address to obtain a list of users interacting with the DEX. Next, we decoded the input values of each transaction using the router contract's application binary interface (ABI), obtaining the corresponding function objects (funcobj) and parameters (funcparam). Based on this, we filtered liquidity providers through keywords in function names. We also established a series of dictionaries to categorize function objects, generating attributes (func_type). Finally, we extracted liquidity data by filtering for keywords "Add Liquidity" or "Remove Liquidity" in function types.

While processing external transactions, we also obtained ERC-20 trading records for the addresses participating in liquidity activities. From the ERC-20 token trading logs, we could identify the amounts of token pairs that liquidity providers gained or forfeited when adjusting liquidity based on transaction hash values, then converted them to USD based on the prices on the transaction day. After the above preprocessing steps, we obtained timestamps and amounts of liquidity changes for each liquidity provider, which could be formatted as time series for further analysis.

As of July 2021, the trading records of Uniswap and SushiSwap router contracts were 46,077,169 and 2,030,355, respectively. Grouped by address, we obtained 2,310,175 and 160,345 unique addresses, of which 297,345 and 43,705 addresses participated in liquidity provision, accounting for 12.8% and 27.2% of the total addresses. By comparing the address lists of the two platforms, we found 50,176 overlapping addresses, of which 27,521 were duplicate liquidity providers.

We categorized the functions called by the Uniswap and SushiSwap contracts based on their naming conventions and functionalities into five types. Among them, three types are "Swap" functions: "ETH-Token," "Token-ETH," and "Token-Token," corresponding to the three trading pairs on the DEX; the other two types are liquidity functions: "Add Liquidity" and "Remove Liquidity."

In the table, # Called indicates the number of times the corresponding function was called, % Called indicates the percentage of the corresponding function out of the total calls, and % Called by Address indicates the percentage of addresses that called the corresponding function out of the total addresses.

Table 1: Number of Calls for Corresponding Functions in Uniswap

and Percentage of Addresses Calling the Corresponding Functions

Table 2: Number of Calls for Corresponding Functions in SushiSwap

and Percentage of Addresses Calling the Corresponding Functions

From the statistics, it is evident that although Uniswap and SushiSwap provide the same services, user behavior differs. The number of users on Uniswap far exceeds that on SushiSwap, but in terms of % Called by Address, the usage of the five function types by users on both platforms is not proportional. Uniswap users are more likely to use "Swap" functions, especially for ETH-token trades. In contrast, SushiSwap users prefer token-token trades, which account for 34.69% of user numbers and contribute 31.59% of trading volume. Additionally, SushiSwap's liquidity functions have higher call counts and user percentages than Uniswap.

Methodologically, we captured user groups with similar behaviors through formatted trading flows, constructed similarity graphs of trading flows, and applied unsupervised clustering algorithms (note: the original text contains a section detailing the methodology, which is omitted here).

6.4 Dataset Analysis Results

Based on time series data and the behavior of liquidity providers, we explored the impact of governance tokens on liquidity mining.

Adding governance token rewards in liquidity mining can significantly increase the total liquidity value and the number of liquidity providers in the short term. However, in the long run, this is not a particularly effective measure.

Liquidity providers providing less liquidity are less active in the DEX, but the most active liquidity providers do not necessarily come from addresses providing the most liquidity in the DEX. Some medium-sized liquidity providers more frequently add and remove liquidity to participate in liquidity mining across multiple protocols and earn governance tokens. In contrast, heavy liquidity providers tend to pursue long-term trading fee income and are less influenced by other external factors.

6.5 Total Value Locked (TVL)

Next, let us analyze the trends in total liquidity value (in USD) for Uniswap and SushiSwap from August 2020 to 2021 in chronological order.

Figure 2: Total Liquidity Value and Daily Trading Volume of Uniswap and SushiSwap

First, regarding the vampire attack. During this phase, SushiSwap's attack caused Uniswap's liquidity to plummet from about $3 billion to nearly $2 billion, further dropping to around $500 million within days, although this figure was still higher than a week prior. However, Uniswap's trading volume remained high at approximately $300-800 million daily. Therefore, we can infer that despite SushiSwap incentivizing liquidity mining through governance tokens, some liquidity providers still had confidence in Uniswap's trading volume and chose to stay on Uniswap to earn commission income.

On September 18, 2020, Uniswap launched its counterattack, and its total liquidity value quickly surpassed that of SushiSwap, maintaining an overwhelming advantage until November 18, 2020, when Uniswap's liquidity mining activities incentivized by UNI governance tokens ended, leading many Uniswap liquidity providers to switch to SushiSwap. The event of UNI token release on September 18 brought Uniswap a total liquidity value of $1.65 billion, while on the same day, SushiSwap lost $159 million in total liquidity value. In the following two months, Uniswap's total liquidity value continued to rise, peaking at $3.06 billion on November 14. Afterward, just three days before the end of the activity, some liquidity providers began to withdraw liquidity, causing a slight decline in total liquidity value. On November 18, Uniswap's total liquidity value plummeted by $1.29 billion, while SushiSwap's total liquidity value increased by $578 million, again resembling a "vampire attack."

After November 18, 2020, the total liquidity values of Uniswap and SushiSwap began to align with the price trends of ETH. Notably, on January 6, 2021, and April 1, 2021, the growth of total liquidity value coincided with significant increases in Ethereum prices. This can be interpreted as the price and exchange rate of ETH, as a primary currency, directly influencing the total liquidity value calculated in USD. From a market perspective, the appreciation of Ethereum may have activated the blockchain market, leading to more frequent token trading. Consequently, liquidity providers could increase their earnings by participating in more trades.

The fluctuations in trading volume do not show a clear correlation with selected events but are more related to ETH prices. However, changes in trading volume also indicate changes in liquidity providers' commission income. Therefore, governance tokens played a crucial role in incentivizing the total liquidity value of SushiSwap, even though its trading volume consistently remained lower than that of Uniswap.

After November 18, 2020, the total liquidity values of Uniswap and SushiSwap began to align with the price fluctuations of ETH. Particularly on January 6, 2021, and April 1, 2021, the increase in total liquidity value coincided with sharp rises in Ethereum prices. This can be interpreted as the price and exchange rate of ETH, as a primary currency, directly influencing the total liquidity value calculated in USD. From a market perspective, the appreciation of Ethereum may have led to a more active blockchain market, resulting in more frequent token trading. Thus, liquidity providers could earn more by participating in more trades.

The trading volume chart indicates that fluctuations in trading volume do not have a particularly precise correlation with selected events but are more related to ETH prices. However, changes in trading volume imply changes in liquidity providers' commission earnings. Therefore, governance tokens played a key role in incentivizing the total liquidity value of SushiSwap, despite its trading volume always being lower than that of Uniswap.

6.6 Types of Called Functions

We obtained time series data for five functions ('ETH-Token', 'Token-ETH', 'Token-Token', 'Add Liquidity', and 'Remove Liquidity') based on daily function call counts.

Figure 3: Function Time Series of Uniswap and SushiSwap

The above figure shows the activity levels of liquidity providers within 3-4 days after the issuance of governance tokens. It is evident that the announcement of SUSHI rewards increased the frequency of function calls on SushiSwap, particularly for the "Add Liquidity" function. However, after the vampire attack, while the "Add Liquidity" function remained high for several days, it was accompanied by an increase in the "Remove Liquidity" function. This may be due to two reasons: first, SUSHI rewards decrease over time, prompting some liquidity providers to withdraw funds; second, Uniswap took countermeasures. This aligns with the competitive strategies of the two platforms in early September.

Additionally, all five functions experienced simultaneous spikes during certain periods, such as the Uniswap heat map in September 2020 and the SushiSwap heat map in January 2021, where all five lines exhibited vertical dark patterns. From the rows, it can be seen that the call counts for swap functions are lower than those for liquidity functions. For Uniswap, the function for swapping tokens for ETH is particularly prominent compared to other functions, reflecting the primary purpose of user interaction with Uniswap.

6.7 Time Distribution of New Users and New Liquidity Providers

Figure 4: New Users and Liquidity Providers of Uniswap and SushiSwap

We obtained timestamps for addresses that first interacted with the DEX and first provided liquidity from the trading records of the two DEXs and plotted the above figure. Although some addresses initially acted as traders before becoming liquidity providers, this does not affect our ability to count the number of new users or new liquidity providers by date and observe their temporal characteristics. Notably, the trading volumes of Uniswap and SushiSwap differ significantly, leading to discrepancies in the vertical axis of the figure, but this does not hinder us from discovering useful information from the trends.

By comparing the curves of new users and new liquidity providers in the figure, we can infer the motivations for users to use Uniswap or SushiSwap at different times.

From the curve of Uniswap, the changes in the two curves are inconsistent, which aligns with Uniswap's high trading volume: most users use Uniswap for trading. At the beginning of September 2020, during the vampire attack, the curve for new liquidity providers exhibited two prominent peaks.

In contrast, the curve for SushiSwap shows that the two curves converged only in January 2021, indicating that most users on SushiSwap aimed to provide liquidity and earn SUSHI rewards rather than trade.

Additionally, we can observe fluctuation patterns in September 2020, November, and January and May 2021. This suggests that certain special events, such as rising Ethereum prices or policy changes, have some impact on user behavior and motivations, attracting more new users. For instance, after January 2021, as ETH prices rose, the daily number of new users on Uniswap and SushiSwap also increased accordingly, peaking in May 2021, before gradually declining and stabilizing. This macro fluctuation related to the blockchain market is not closely related to the services of the DEX itself and user interests.

6.8 Clustering Results Analysis

We classified liquidity providers into six categories based on their on-chain activities to analyze their interaction purposes with the DEX. We performed clustering for both Uniswap and SushiSwap.

Optional liquidity providers refer to those who provide minimal liquidity and lack active participation. In Uniswap and SushiSwap, they account for 58.1% and 17.2%, respectively. Their total locked value is very low (less than $1,000), and their operational intervals are significant. We speculate that these addresses merely want to try liquidity mining without committing substantial funds.

Light liquidity providers refer to those who provide more liquidity. In Uniswap and SushiSwap, they account for 18.0% and 42.2%, respectively. Light liquidity providers can be further divided into two sub-clusters based on operational frequency: inactive light liquidity providers and active light liquidity providers. Active light liquidity providers interact more frequently with the DEX, particularly on Uniswap. Although the ratios of light liquidity providers and optional liquidity providers differ between the two DEXs, their combined ratio exceeds 59%, indicating that liquidity providers offering minimal liquidity constitute the vast majority in both DEXs. Furthermore, the higher proportion of light liquidity providers on SushiSwap suggests that long-term governance token rewards can attract more small-scale users to participate in liquidity mining.

Medium liquidity providers refer to those who provide relatively high liquidity. In Uniswap and SushiSwap, they account for 19.9% and 29.7%, respectively. Medium liquidity providers can be further subdivided into risk-seeking medium liquidity providers and risk-averse medium liquidity providers. The liquidity distribution values of both types of medium liquidity providers are similar, but their operational frequencies differ. Specifically, risk-averse medium liquidity providers are more cautious or passive due to concerns about potential losses. In contrast, risk-seeking medium liquidity providers prefer to provide liquidity across multiple DEXs to earn governance token rewards with high annual percentage yields, even if they may face impermanent losses or risks of capital theft.

Heavy liquidity providers refer to those who provide substantial liquidity. They represent the lowest proportion in both DEXs, accounting for 4.6% and 11%, respectively. They have significant liquidity scales in the DEX, for example, address 0xf0fc provided over $14 million USDC and 7,861 ETH on Uniswap, worth nearly $30 million at the time. Additionally, compared to risk-seeking medium liquidity providers, these addresses have lower operational frequencies, indicating that they tend to pursue long-term commission income, and their decisions are less influenced by other external factors.

6.9 Behavior of Liquidity Providers in Uniswap

In this section, we collected a list of Uniswap liquidity provider addresses that operated liquidity or called special contracts during a specific period. This period started on August 28 and ended on November 18, when Uniswap ceased UNI token incentives.

Due to SushiSwap launching the Masterchef contract, allowing users to stake UNI-V2 tokens to earn SUSHI, many new addresses flocked to Uniswap to provide liquidity in order to obtain UNI-V2 tokens. In just ten days, the incremental number of liquidity providers on Uniswap was nearly ten times the total of the previous three months. Among these new liquidity providers, the proportion of heavy and medium liquidity providers was significantly higher than that of returning liquidity providers. We found that during the period from August 28 to September 8, over half of the liquidity providers staked their UNI-V2 tokens with MasterChef to earn more SUSHI, while less than 5% of returning liquidity providers did so. We speculate that this is because optional liquidity providers account for a higher proportion among returning liquidity providers, and these liquidity providers are less active and provide less liquidity. Therefore, they may be less sensitive to market information and less concerned about SUSHI, as their capital investment is minimal, preventing them from earning substantial SUSHI.

In the subsequent phase, we noticed that nearly half of the addresses earning SUSHI withdrew liquidity within the next two months. Based on the timing of withdrawals, we divided them into two periods: September 9 to September 17 (Period A) and September 18 to November 18 (Period B). First, 26.7% of liquidity providers chose to withdraw liquidity during Period A, of which 57.7% were medium and heavy liquidity providers. We believe that this group of liquidity providers is the most sensitive to the market and values liquidity income the most, providing liquidity only during periods of higher yields. Second, 21.2% of addresses chose to withdraw liquidity from SushiSwap during Period B. We speculate that this is because Uniswap released UNI on September 18 and initiated liquidity mining, quickly attracting back some liquidity providers who had exited from SushiSwap due to the high daily trading volume and trading fee income brought by Uniswap. Finally, after the cessation of UNI issuance on November 18, only a very small number of addresses in this group migrated liquidity to SushiSwap by calling the Migrator contract.

6.10 Overlapping Addresses

Uniswap and SushiSwap had 297,345 and 43,705 independent addresses that provided liquidity, respectively, with 27,521 addresses providing liquidity to both exchanges. This section will analyze the clustering characteristics of these overlapping addresses.

Figure 5: Short-term and Long-term Changes of Overlapping Addresses in SushiSwap Clusters from September 8, 2020, to November 18, 2020, and from September 8, 2020, to July 18, 2021

The data shows the short-term and long-term changes of overlapping addresses in SushiSwap clusters. In early September 2020, SushiSwap launched the vampire attack, but this advantage quickly disappeared after Uniswap initiated liquidity mining and UNI. By November 2020, we can see significant overlap between Uniswap and SushiSwap liquidity providers during the first two months of SushiSwap. Specifically, 87.2% of SushiSwap's heavy liquidity providers had overlapping addresses, with 76% classified as medium and heavy liquidity providers of Uniswap. However, nearly a year after the launch of SushiSwap, data from July 18, 2021, shows that the proportion of overlapping addresses decreased, with the overlap rate of SushiSwap's heavy liquidity providers dropping by 41.4%. Therefore, we can conclude that SushiSwap gradually cultivated a dedicated group of stable liquidity providers through long-term SUSHI incentives over nearly a year.

7. Conclusion

We confirm that adding governance tokens in liquidity mining has different attractions for different types of liquidity providers.

Based on clustering results, using SushiSwap as an example, we found that over 50% of heavy and medium liquidity providers withdraw funds in a short time. We summarize two reasons behind this phenomenon:

1) High rewards from competitors (Uniswap);

2) The annualized yield of rewards decreases over time.

Thus, rewarding with governance tokens in the early stages of the protocol is ineffective.

We found that the most active participants in liquidity mining are not those with the most funds, but rather addresses with moderate funds. The reason for this phenomenon is that this group of addresses is subjectively the most motivated to pursue benefits and rewards, leading us to infer that incentives like governance tokens should have the greatest appeal to them.

From a broader perspective, all these concerns regarding the impact of governance tokens indicate that they have not fulfilled their designed governance value but have been used as arbitrage tools by relatively well-funded speculators, failing to positively influence the development of the protocol. However, we also note that the neglect of governance capabilities has not diminished their appeal to users. By comparing the overlapping liquidity provider ratios of SushiSwap and Uniswap in two phases, we find that SushiSwap has continuously expanded its stable user base and scale through long-term governance token issuance and business innovation.

8. Future

Protocols often attract users by adding governance tokens in liquidity mining. Our analysis results show that while this method can attract more users in the short term, it is challenging to retain them. Therefore, protocols should seek other ways to attract users, such as reducing trading costs and providing more convenient features.

Moreover, governance tokens are the foundation of community autonomy and an essential component of any Web 3.0 protocol. As a pioneer in distributing governance tokens, liquidity mining can only encourage liquidity providers to participate passively rather than contribute actively. We realize that higher participation thresholds, such as rising gas fees and sufficient liquidity, hinder liquidity providers with insufficient funds. Therefore, exploring new governance token distribution methods is crucial for the Web 3.0 community. Decentralized communities need to explore new incentive mechanisms that can actively motivate members to participate in governance. For example, participants could propose suggestions related to the development of the protocol, and if the community adopts these suggestions, the proposers would receive governance token rewards. Future research can explore, model, and evaluate optimized governance mechanisms to build better decentralized communities.