This analysis aims to discern the proportion of trades on various DEXes driven by arbitrage or MEV-using participants. By understanding the levels of user sophistication, we can gain insights into the degrees of automation and profit-seeking behavior in a DEX's user base. This, in turn, reveals their responsiveness to changes in incentives and costs.
Previously, research into the dynamics of DEX trading has been conducted, offering insights into the composition of volume across different user types. Some of this research painted a picture of an ecosystem majorly driven almost entirely by opportunistic MEV bots (~80%), with retail users playing a minimal role. By using more granular user attribute labels, our current research delves deeper and paints a more optimistic scenario regarding user diversity on DEXes.
Before diving deep into the analysis, it's crucial to understand a few fundamental concepts that play pivotal roles in the DEX trading landscape.
Liquidity Providers (LPs)
Liquidity providers are crucial participants in decentralized exchanges. They deposit assets into liquidity pools, enabling traders to swap between different tokens without the need for a counterparty. In return for their service, LPs earn fees from the trades that pass through the pool. The mechanics and incentives for liquidity provision have been extensively explored in Uniswap's whitepaper.
Maximal Extractable Value (MEV)
MEV refers to the value miners can extract from reordering, including, or censoring transactions. In the context of DEXes, traders can offer miners a bribe to, for example, place their trade before others in a block, ensuring they get the best price. This phenomenon becomes especially important during times of network congestion or particularly lucrative trading opportunities. Flashbots, an organization researching MEV strategies, provides a comprehensive overview of its effects on Ethereum's ecosystem.
Sandwich attacks are a type of MEV strategy, where a trader (usually a bot) notices a large incoming trade that will move the market. They place a trade before (the 'front-run') and after (the 'back-run') the large trade, benefiting from the price slippage the large trade causes. It's named a 'sandwich' because the large trade gets sandwiched between the attacker's two trades. The speed and precision required to execute such a strategy suggest a high degree of automation.
Onchain arbitrage refers to the practice of taking advantage of price discrepancies between tokens on the same blockchain, all executed in a single transaction (fully-routed). Given the open and transparent nature of blockchain data, these opportunities are quickly spotted and acted upon, often by automated bots. The speed at which these arbitrage opportunities are exploited often requires the use of automated tools and scripts. The Flash Boys 2.0 paper dives deep into the mechanics and challenges of onchain arbitrage in a decentralized setting. While there are other forms of arbitrage that involve different markets, particularly those off-chain on centralized exchanges, this analysis focuses only on fully-routed onchain arbitrage transactions.
The above concepts underscore the evolving landscape of DEX trading, characterized by sophisticated strategies and rapid execution. This high-speed, automated environment directly influences the dynamics of liquidity and trading costs. The agility of automated players, be it arbitrageurs or those leveraging MEV, sets the tone for the entire ecosystem's reaction to changes in DEX policies or incentives.
- Liquidity Mining Incentives: Automated players will likely adjust swiftly following the removal of incentives, especially if Liquidity Provider (LP) fees aren't sustainably increased. In comparison, human participants might exhibit a delayed response. Cohort analysis can help plan better incentive programs that achieve greater efficiency and impact.
- Protocol Fees: Automated and short-term traders might be sensitive to any rise in trading costs. However, for platforms like Uniswap X, if superior price execution is consistently demonstrated compared to alternatives, these players might still find it favorable to trade despite the fee. Cohort analysis can help better understand the pros and cons of potential fee adjustments.
User attribute definitions
Arbitrage Flagging: We identify arbitrage trades based on specific criteria. Note that we focus exclusively on fully-routed, onchain arbitrage. Our definitions align with Flashbots:
- Input token must equal output token
- contract_address must match the from_address (sender)
- OR contract_address must match the to_address
- OR from_address (sender) must match to the to_address
- A profitable difference between input and output token quantities.
Aggregator Flagging: Transactions originating from routing contract addresses of platforms. We’ve flagged contract addresses belong to the following aggregators:
- CoW Swap
LP Flagging: Addresses that have previously contributed liquidity to SushiSwap or Uniswap.
Sandwicher Flagging: Employing external Flashbots' data, we flag users (not transactions) who executed sandwich attacks between 2022-11-02 and 2023-03-24.
MEV Flagging: This employs the latest Nansen address flagging data. The factors used in identifying MEV include:
- frequent and high quantity trades
- unverified closed source smart contracts
- complex transactions involving two or more DEXes and/or bridges or two or more transactions within the same block
- abnormal gas fees
Smart Contracts: This includes contracts responsible for routing, other pools, or other infrastructure-related tasks.
Direct traders: This cohort comprises what we conventionally consider as "retail" users. It encompasses any user not fitting into the above categories and are unlikely to be automated.
Address statistics and attributes were sourced from Flashbots and Nansen, whereas DEX swap data was sourced onchain. The analysis is limited to activities on Ethereum Mainnet for the following DEXes: Uniswap V2/V3, SushiSwap, Balancer and Curve. Volume information was calculated using the price of tokens with the highest available liquidity at the time of trade, unless the value of the trade exceeded the total liquidity of that token.
Segmenting trading volume based on user attributes across the four major DEXes we examined shows that a 40-60% share of the volume originates from sophisticated users employing automated methods in their transactions. This includes MEV bots, arbitrageurs, and users who have previously executed a sandwich attack. Given the intricate nature of these strategies, especially the need for real-time data analysis and immediate response, automation becomes not just beneficial but almost essential:
The data visualization paints a vivid picture of the trading landscape across decentralized exchanges, underscoring a consistent theme: automated users are the driving force behind the bulk of trading volume. Their dominance is not merely marginal; they outstrip all other user groups by a wide margin, reinforcing the sophisticated nature of the current decentralized trading environment. This cohort of automated users consists of users that deploy sophisticated methods of trading, such as MEV and sandwich attacks or fully-routed arbitrage.
Diving deeper into the data, it's evident that the influence of automated traders is not uniform across all liquidity tiers. In mid-cap liquidity pools, for instance, they are responsible for a staggering 50% to 70% of the trading volume. Such pools often offer the perfect balance for these traders – sufficient liquidity to ensure meaningful profits, yet not so vast that individual trades have a negligible impact.
Nevertheless, it's essential to note that the majority of trading volume still lies in the highest TVL pools. Uniswap, for example, sees a concentration of its trading volume in its top-tier liquidity pools, constituting 64% of the total volume on the platform:
Pivoting slightly, it's worth noting another group that plays a pivotal role in shaping the DEX landscape – the "aggregators.” While aggregators are considered a separate user cohort, alone they account for 6-10% of volume on Uniswap and higher proportions on other DEXes. While aggregators are motivationally different from arbitrageurs and MEV bots, their routing logic should, in theory, be driven to automatically complete trades on DEX platforms that offer the best price execution. This point is important in that if additional fees are levied against traders of a particular DEX, then the aggregator should take this into account when formulating routing decisions.
The dominant presence of automated volume on DEXes underscores a noteworthy point: the majority of DEX traders operate with a high degree of sophistication, employing complex strategies and advanced tools. This has profound ramifications for any initiatives aiming to enhance rewards or impose additional trade-related costs. It's plausible to infer that these adept traders will promptly evaluate and adapt to such shifts. They'll likely measure the merits of liquidity incentives or temporary bonuses against the profitability landscape, ensuring they opt for the most lucrative. Such evaluations might also span beyond Ethereum mainnet to encompass other networks. Conversely, adding protocol fees or other cost-increasing measures could trigger an immediate reaction unless the DEX offers advantages (like superior price execution) that counterbalance these added expenses.
Miscellaneous Smart Contracts
Following closely are the "Miscellaneous smart contracts" cohort. These represent swaps that journey through various DEX pools, routers or are initiated by a smart contract rather than an individual’s address. Some examples include Uniswap: V2 Router 2, SushiSwap: Router and Curve fi DAI/USDC/USDT pool. These smart contracts contribute significantly to the efficiency and liquidity of the DEX’s ecosystem. Whether through routing agents optimizing trade execution or institutional firms deploying bespoke contracts, these actors collectively contribute a significant amount of volume and liquidity to decentralized exchanges. It is essential to recognize the diverse nature of contracts within this cohort, underscoring the ongoing need for effective wallet tagging at scale to facilitate transparent tracking and analysis.
In the case of Uniswap, since 2023-01-01, the following miscellaneous smart contracts have been the top 5 addresses in terms of volume, with similar patterns found in other DEXes:
The entities within this category range from automated MEV helpers to intricate routers that optimize trade execution across different liquidity pools. As illustrated by the dominant addresses in Uniswap's trade volume since the start of 2023, the vast amounts transacted by these entities underscore their significance.
Direct Trader Volume: Retail Users
Users from the direct trader cohort account for almost 25% of Uniswap’s volume - however, there are further distinct entities within this group. Examining the top contributors to volume in the last year, we can identify addresses belonging to Wintermute Trading collectively contributing approximately 25% of the volume in the direct trader cohort.
We have opted to exclude Wintermute’s volume contribution from the next two charts so as to better understand what are typical order sizes and contributions of non-institutional users; retail users.
Manual/retail user (any user that isn’t a contract address or an automated entity that is interacting directly with the protocol) volume share varies by DEXes - Uniswap having the highest proportion of Retail traders relative to the platform’s total volume is unsurprising given Uniswap’s market share dominance (58% of volume across all networks).
The accompanying chart showcases an intriguing distribution. A staggering 62.3% of the Retail cohort users traded a total volume of up to just $1k in 2023. However, in the past year over 63% of the cohort’s volume from these subgroups came from users with a total volume of between $100k-$500k. Also interestingly, 11% of the cohort’s volume is accounted for by 9 distinct, elite DEX trading individuals. These figures provide a window into the skewed nature of individual traders and how few individuals contribute a significant portion of DEX volume.
Additionally, while the retail cohort makes up approximately 15% of volume, over 70% of Uniswap’s user base belongs to this retail cohort, emphasizing its pivotal role in the ecosystem.
In the context of other DEXes seeing low Retail volume make-up, this could be due to the accounting of volume attributed to routing contracts from trades originating on other DEXes. For instance, a user may initiate a trade on Uniswap but through subsequent steps routing contracts (categorized as a miscellaneous contract) will route volume to other DEXes wherever there is favorable liquidity and price execution. The outcome is that the swap that occurs on the other DEX is attributed to the smart contract and not the Retail user who initiated it. The volume share attributed to Retail more closely approximates users who’ve initiated a trade on that respective DEX. The table below estimates the percentage of transactions since the start of 2023 have interacted with another DEX prior to reaching each respective DEX.
The estimate is one of a lower-bound, where only the four mentioned DEXes were considered. It is highly likely that these percentages are higher if considering other DEXes (e.g., if a transaction first had touched PancakeSwap prior to reaching Uniswap).
The aim of this analysis was to assess the extent to which users on different DEXes, with a specific focus on Uniswap, could be identified as actively engaging through automation. Our findings revealed that nearly half of the trading volume could be attributed to automated entities, including MEV bots, arbitrageurs, and addresses previously involved in sandwich attacks. Encouragingly, Uniswap maintains a notably high proportion of human users compared to other DEXes, which can likely be attributed to its strong brand presence and significant market share in the decentralized trading landscape.
The decentralized trading landscape has emerged as a sophisticated ecosystem, with various user segments contributing to its dynamism. This study aimed to delve deeper into these segments, understanding their distinct behaviors and their individual contributions to DEX trading volumes. The findings underscore the dominance of automated participants, highlighting a core aspect of the DEX environment: the extent of sophistication and, consequently, the propensity for automation among traders.
Several concrete implications arise from this analysis:
- Strategic incentives matter: Given the dominance of automated participants, DEX platforms must carefully strategize their fee structures and liquidity incentives. Even minor changes could lead to swift adjustments in trading behaviors, with potential cascading effects on volume and liquidity. Setting fees haphazardly without justifiable price execution benefits will alienate these sophisticated traders, potentially resulting in a sharp decline in trading volume. This could, in turn, decrease the liquidity of the platform, making it less attractive for both automated and retail traders.
- Conversely, incentive programs that are not carefully calibrated run the risk of being ineffectual, or even counterproductive. For instance, overly generous rewards might draw in liquidity providers in the short term, but if these rewards are unsustainable, they can result in a rapid exodus once they're reduced or removed. Additionally, improperly structured incentives might be exploited by savvy automated participants, leading to a drain on incentive spend without delivering the desired increase in genuine trade activity. Thus, DEX platforms must strike a balance, ensuring that incentives foster long-term growth and stability while safeguarding against opportunistic behaviors.
- Retail Engagement: While automated entities dominate in volume, the vast majority of DEX users fall within the retail cohort. Platforms must ensure a user-friendly experience for these individuals, as they represent a significant portion of the ecosystem's user base and can influence platform reputation.
- Though platforms might be inclined to favor institutional or advanced traders by offering them superior price execution, this shouldn't come at the expense of retail users. Any approach that seems to unduly direct prime orders to institutional players, while relegating retail traders to less favorable trades, can lead to a diminished trust in the platform. A perception of preferential treatment can deter retail users from engaging, believing they're at a disadvantage. Over time, this can erode a platform's user base, resulting in reduced liquidity, lower trading volumes, and a potential tarnishing of the platform's brand. It's essential for DEX platforms to maintain an equitable trading environment, ensuring that all users, irrespective of their sophistication or trading volume, feel they're receiving a fair and transparent service.
In summary, the challenge lies in harmonizing the advanced requirements and sensitivities of automated traders with the usability needs of individual users. Especially for leading platforms like Uniswap, this equilibrium is crucial. Adjustments to fee structures or liquidity incentives can have immediate repercussions due to the agility of automated participants. Simultaneously, ensuring a positive experience for the vast retail cohort is essential for maintaining platform reputation and trust.
In essence, the future of DEX platforms hinges on their ability to cater to this dual user base, striking a balance that ensures stability, growth, and continued relevance in the decentralized trading sphere.
At Gauntlet, we're deeply invested in these dynamics. This analysis not only validates our assumptions about users but also guides our strategic deployment of incentive programs for Uniswap. As we move forward, these insights help us better understand the impact of fees and incentives, ensuring maximum benefit for the protocol.
Recommended areas of future research
The analysis here focused only on user cohort and automation aspects within the onchain domain. Understanding the symbiotic relationship between decentralized and centralized exchanges can reveal the nuances of liquidity flow, trade execution speeds, and price disparities. This encompasses not just the technical aspects but also the behavioral tendencies of traders as they navigate between these two markets.
Additionally, arbitrage opportunities that span across both centralized and decentralized platforms might be more common than currently perceived. A deep dive into these strategies can unravel the complexities of cross-platform trades. This would involve studying the latency in trade executions, the cost implications, and the potential barriers or facilitators to such arbitrage opportunities.
Finally, gaining insights into arbitrage opportunities that span multiple blockchains can shed light on patterns of liquidity migration. When assets transition between chains, how does this influence arbitrage dynamics? Moreover, do certain blockchains inherently present more favorable conditions for automated traders?
This analysis offers valuable insights into the user composition of various DEXes. However, several limitations must be acknowledged to fully appreciate the context and potential areas for improvement. By highlighting these limitations, we aim to underscore areas that could be expanded upon in future research endeavors.
1. Sandwich Attack Data:
Our dataset for sandwich attacks was confined to a specific and narrow date range earlier this year. Consequently, our method for identifying "sandwichers" was based on whether an address had ever executed such an attack, rather than flagging specific transactions as sandwich attacks. This approach was adopted as our primary interest lay in understanding the overall user cohort composition of DEXes, rather than precisely attributing volume to sandwich attacks.
2. Liquidity Provider Data:
Our insights into Liquidity Providers (LPs) were based solely on data from SushiSwap and Uniswap, omitting potential contributions from other platforms. Future analyses would benefit from incorporating a wider array of LP data from various other DEXes.
3. Address Labeling Challenges:
Effectively categorizing wallet addresses, especially at a large scale, presented challenges. In subsequent analyses, we aim to differentiate institutional actors from other user cohorts, providing a clearer picture of market participants.
4. Arbitrage Scope:
The study focused exclusively on onchain arbitrage, neglecting potential opportunities that involve centralized platforms. Exploring arbitrage activities that span both decentralized and centralized exchanges would be enlightening. Such a study could reveal price efficiency disparities between DEXes and CEXes and determine if these arbitrage opportunities are more prevalent than other types.
5. Volume Accounting:
The common approach to calculating volume on a decentralized exchange involves aggregating the volumes of individual swaps within a specific transaction hash. Such a calculation can result in an amplification effect where by a majority of volume can be attributed to other processes by way of complexity: as more swaps occur within a single transaction, the initiator of the trade is responsible for a smaller proportion of the total volume. However, it's important to note that onchain swaps are interconnected across different DEXes. Simplifying swaps into a single-row summary can potentially overlook the contributions from other liquidity sources.