Today, our research team published a study on ethresear.ch, delving into the impact of latency (time) on MEV extraction. More specifically, we demonstrate the costs associated with introducing artificial latency within a PBS (Proposer-Builder Separation) framework. Additionally, we present findings from Adagio, an empirical study that explores the implications of latency optimization aimed at maximizing MEV capture.
In late August 2023, we launched Adagio, a latency-optimized setup on the Ethereum mainnet. The primary objective was to collect actionable data ethically, with minimal disruptions to the network. Until this point, Adagio has not been a client-facing product, but an internal research initiative running on approximately 100 self-funded validators. We initially shared ongoing results of the Adagio pilot in our Q3 Quarterly Insights report in October.
In alignment with our commitment to operational honesty and rational competition, this study discloses the full results of Adagio, alongside an extensive discussion of node operator incentives and potential adverse knock-on effects on the Ethereum network. As pioneers in MEV research, our primary objective is to address and mitigate existing competitive dynamics by offering a detailed analysis backed by proprietary data from our study, which will be explored further in the subsequent sections of this article.
This article offers a top-level summary of our study, contextualizing it within the ongoing Ethereum community dialogue on ethically optimizing MEV performance. We dive into the key findings of the study, highlighting significant observations and results. Central to our discussion is the exploration of the outcomes tied to the implementation of the Adagio setup, which demonstrates an overarching boost in MEV capture.
Ultimately, we recognise that node operators are compelled and incentivised to employ latency optimization as a matter of strategic necessity. As more operators take advantage of this inefficiency, they set a higher standard for returns, making it easier for investors to choose setups that use latency optimization.
This creates a cycle where the use of latency optimization becomes a standard practice, putting pressure on operators who are hesitant to join in. In the end, the competitive advantage of a node operator is determined by their willingness to exploit this systematic inefficiency in the system.
Additionally, we demonstrate that the parameters set by our Adagio setup corresponds to an Annual Percentage Rate (APR) that is 1.58% higher than the vanilla (standard) case, with a range from 1.30% to 3.09%. Insights into these parameters are provided below, with additional clarity available in the original post.
Let’s preface this section with the phrase - Right Place at the Right Time.
Delightfully analogous to the quote above, we’re adding further insights to the overarching discourse on the implication of latency optimization (i.e, a strategy where block proposers intentionally delay the publication of their block for as long as possible to maximize MEV capture) when it has become a burning topic within the Ethereum community, drawing increased attention from various stakeholders concerned about its network implications.
Yet, despite its growing significance, there has been a noticeable lack of empirical research on this subject. As pioneers in MEV research, we've been investigating this concept for over a year, incorporating latency optimization as one of our MEV strategies from the outset. Now, we're proud to contribute to the ongoing discussions and scrutinize the most significant claims with robust, evidence-based research.
In a previous article about Chorus One’s approach to MEV, we emphasized the importance of exploring the dynamics between builders, relays, and validators with the dimension of time.
Our focus on how latency optimization can profoundly influence MEV performance remains unchanged. However, we've identified a crucial gap in empirical data supporting this concept. Compounding this issue, various actors have advocated for methods to increase MEV extraction without rigorous analysis, resulting in inflated values based on biased assumptions. Recognizing the serious consequences this scenario poses in terms of centralization pressure, we now find it imperative to conduct a deep dive into this complex scenario.
Our strategy involves implementing a setup tailored to collect actionable data through self-funded validators in an ethical manner, ensuring minimal disruptions to the network. This initiative is geared toward addressing the existing gap in empirical research and offering a more nuanced understanding of the implications of latency optimization in the MEV domain.
The key objectives of this research is three-fold, including:
In the following section, we will present a comprehensive overview of the three most pivotal and relevant observations from the study, and as promised earlier, we will also delve into the results of Adagio.
Context: First, we delve into PBS inefficiencies and MEV returns.
Here, we explore the inefficiencies in the Proposer-Builder Separation (PBS) framework, showing how timing in auctions can be strategically exploited to generate consistent, excess MEV returns.
Additionally, we demonstrate how all client-facing node operators are incentivized to compete for latency-optimized MEV capture, irrespective of their voting power.
Key Finding: Latency optimization is beneficial for all client-facing node operators, irrespective of their size or voting power.
Using an empirical framework to estimate the potential yearly excess returns for validators who optimize for latency considering factors like the frequency of MEV opportunities, network conditions, and different latency strategies, our results indicate that node operators with different voting powers have varying levels of predictability in their MEV increases.
The above figure demonstrates that higher voting power tends to result in more predictable returns, while lower voting power introduces more variance. The median weekly MEV reward increase is around 5.47% for a node operator with 13% voting power and 5.11% for a node operator with 1% voting power.
The implication here is that big and small node operators cater to different utilities of their clients (delegators) because they operate at different levels of risk and reward. As a result, optimizing for latency is beneficial for both small and large node operators. In simpler terms, regardless of their size, node operators could consider optimizing latency to better serve their clients and enhance their overall performance.
As we look at a longer timeframe, the variability in rewards for any voting power profile is expected to decrease due to statistical principles. This means that rewards are likely to cluster around the 5% mark, regardless of the size of the node operator.
In practical terms, if execution layer rewards make up 30% of the total rewards, adopting a latency-aware strategy can boost the Annual Percentage Rate (APR) from 4.2% to 4.27%. This represents a noteworthy 1.67% increase in overall APR. Therefore, this presents a significant opportunity, encouraging node operators to adopt strategies that consider and optimize for latency.
Context: Second, we discuss the costs of introducing artificial delays, explaining how it increases MEV rewards but at the expense of subsequent proposers.
Key Finding: MEV tends to benefit node operators with higher voting power, giving them more stable returns. When these operators engage in strategic latency tactics, it can increase centralization risks and potentially raise gas cost and faster burnt ETH for the next proposer..
While sophisticated validators benefit from optimized MEV capture with artificial latency, the broader impact results in increased gas costs and a faster burning of ETH for the next proposers. The Ethereum network aims to maximize decentralization by encouraging hobbyists to run validators, but the outlined risks disproportionately affect solo validators. Below, we demonstrate that these downside risks are significant in scale, and disproportionately impact solo validators.
Figure 2 illustrates that introducing artificial latency increases the percentage of ETH burned, potentially reducing final rewards. Even a small increase in burnt ETH can significantly decrease rewards, especially for smaller node operators who are chosen less frequently to propose blocks. The negative impact is most significant for solo validators, making them less competitive on overall APR and subject to greater income variability. Large node operators playing timing games benefit from comparatively higher APR at lower variance to the detriment of other operators.
MEV tends to benefit node operators with higher voting power, giving them more stable returns. When these operators engage in strategic latency tactics, it can increase centralization risks and potentially raise gas fees for the entire Ethereum network. Moreover, larger node operators, due to their size, have access to more data, giving them an edge in testing strategies and optimizing latency.
In this scenario, node operators find it necessary to optimize for latency to stay competitive. As more operators adopt these strategies, it becomes a standard practice, creating a cycle where those hesitant to participate face increasing pressure. This results in an environment where a node operator's success is tied to its willingness to exploit systematic inefficiencies in the process.
Context: In late August 2023, Chorus One launched a latency-optimized setup — internally dubbed Adagio — on Ethereum mainnet.
Its goal was to gather actionable data in a sane manner, minimizing any potential disruptions to the network. Until this point, Adagio has not been a client-facing product, but an internal research initiative running on approximately 100 self-funded validators. We are committed to both operational honesty and rational competition, and therefore disclose our findings via this study.
In simple terms, this section analyzes the outcomes of our Adagio pilot, focusing on how different relay configurations affect the timing of bid selection and eligibility in the MEV-Boost auction.
Our pilot comprises four distinct setups, each representing a variable (i.e. a relay) in our experiment:The Benchmark Setup, The Aggressive Setup, The Normal Setup, and the Moderate Setup.
Key Findings: The results of this pilot indicate that the timing strategies opted by node operators used within relay operations have a significant impact on how competitive they are.
The aggressive setup, in particular, allows non-optimistic relays to perform similarly to optimistic ones. This means that certain relays can only effectively compete if they introduce an artificial delay.
In extreme cases, a relay might not be competitive on its own, but because it captures exclusive order flow, node operators might intentionally introduce an artificial delay when querying it or might choose not to use it at all. Essentially, these timing strategies play a crucial role in determining how relays can effectively participate and compete in the overall system.
These results offer valuable insights into how strategically introducing latency within the relay infrastructure can impact the overall effectiveness and competition in the MEV-Boost auction. The goal is to level the playing field among different relays by customizing their latency parameters.
The above graph displays the eligibility time of winning bids in the Adagio pilot compared to the broader network distribution. As expected, Adagio selects bids that become eligible later with respect to the network distribution. Notably, our setup always selects bids eligible before 1s, reducing the risks of missed slots and increased number of forks for the network.
Finally, it’s worth mentioning that our results indicate that certain setups are more favorable to winning bids. This opens up the possibility for relays adopting latency optimization to impact their submission rate.
Bringing together the data on latency optimization payoff and the results of our Adagio pilot allows us to quantify the expected annual increase of validator-side MEV returns.
The simulation results presented in Fig. 4 show that, on average, there is a 4.75% increase in MEV extracted per block, with a range from 3.92% to 9.27%. This corresponds to an Annual Percentage Rate (APR) that is 1.58% higher than the vanilla (standard) case, with a range from 1.30% to 3.09%.
The increased variability in the range is mainly due to the limited voting power in the pilot, but some of it is also caused by fluctuations in bid eligibility times. The observed median value is 5% lower than the theoretically projected value. To address this difference, the approach will be updated to minimize variance in bid selections and keep eligibility times below the 950ms threshold.
Let’s take a moment to consolidate the key takeaways derived from our study and the Adagio setup.
Since inception, Chorus One has recognised the importance of MEV and spearheaded the exploration of the concept within the industry. From establishing robust MEV policies and strategies, receiving a grant from dYdX for investigating MEV in the context of the dYdX Chain to conducting empirical studies that investigate the practical implications of factors influencing MEV returns, we've consistently taken a pioneering role. Our dedication revolves around enhancing the general understanding of MEV through rational, honest, and practical methods.
For comprehensive details about our MEV policies, work, and achievements, please visit our MEV page.
If you’d like to learn more, have questions, or would like to get in touch with our research team, please reach out to us at research@chorus.one.
If you want to learn more about our staking services, or would like to get started, please reach out at staking@chorus.one
About Chorus One
Chorus One is one of the biggest institutional staking providers globally operating infrastructure for 45+ Proof-of-Stake networks including Ethereum, Cosmos, Solana, Avalanche, and Near amongst others. Since 2018, we have been at the forefront of the PoS industry and now offer easy enterprise-grade staking solutions, industry-leading research, and also invest in some of the most cutting-edge protocols through Chorus Ventures.
