A Ray of Light in the Ethereum Dark Forest: Understanding the Game Under the MEV Competition

This article was published on ChainNews, author: Cao Yixin, employed at HashKey Capital Research.

The March report released by the MEV research organization Flashbots states that its third-party Ethereum client version MEV-geth has been adopted by 12 mining pools that collectively hold 58.8% of the network's hash power. The rising popularity of this version coincides with a significant decrease in Ethereum Gas fees. Although the causal relationship has yet to be empirically proven, it is enough to attract widespread attention, and there are many debates within the community regarding the legitimacy of MEV.

This article aims to clarify the definition of MEV and its impact on the Ethereum ecosystem by analyzing the sources of MEV, and to explore a potential state that may form based on the game theory of stakeholders under the MEV interest distribution market order that Flashbots is attempting to establish.

Reference reading:

"Flashbots Transparency Report --- March 2021"

Sources of MEV

Ethereum has matured over the years, forming a stable pattern from producers (miners) to auction houses (mining pools) to bidders (users), regulating the supply and demand of block space through the Priority Gas Auction (PGA) mechanism and block rewards.

If transactions can be divided into three processes: price discovery, transaction matching, and transaction execution, then at this stage, the core value of block space is guaranteed by a series of blockchain core "components" such as hash pointers, consensus mechanisms, and peer-to-peer networks, providing decentralized transfer services (the trustless nature of transaction execution). Furthermore, smart contracts further replace the credit intermediaries relied upon before the transaction was completed (the trustless nature of transaction matching), thereby eliminating the friction costs associated with "trust" throughout the transaction process, providing critical infrastructure for the issuance and trading of digital assets.

Reference reading:

"In-depth Analysis of Ethereum's Block Space Market: New Models in Mining Cycles"

Driven by smart contracts, the Ethereum application layer has spawned large-scale innovative applications in DeFi, causing block space to be filled with complex trading logic. The DeFi system, built from Lego blocks such as over-collateralized lending, AMM, yield farming strategies, derivatives, insurance, and algorithmic stablecoins, has begun to take shape. The wealth effects brought by its composability have been a topic of discussion, while the transparency and openness of block space and the mempool have brought new opportunities and challenges to participants in the price discovery process, with MEV being a byproduct of this characteristic.

Price discovery is an indispensable economic function for the self-organizing development of decentralized markets. Ordinary trading users, professional arbitrage and liquidation bots, and even hackers who discover economic system vulnerabilities are all participants in price discovery (collectively referred to as seekers). They search for the gap between actual and expected outcomes, incurring capital costs and transaction fees to obtain potential profits. The transparency and accessibility of information on the blockchain provide new opportunities for an efficient price discovery system, while innovative lending models like flash loans bring many conveniences to price discovery.

However, these seekers face new challenges. On one hand, they compete in PGA competition for better transaction timing; on the other hand, they face the scrutiny and plundering of front-runners. The competition among seekers highlights and elevates the time value of transactions within blocks, which manifests as higher fee costs than ordinary transactions, and this is understandable. The discreteness of block generation time and the transparency of the mempool allow front-runners to easily monitor the mempool for exploitable transactions without incurring search costs, participating in the PGA process by further raising Gas fees to execute their transactions and those of others in a specific time order, achieving their goal of profit without effort, thereby harming the interests of seekers. Therefore, it is advisable to differentiate the nature of these two types of transactions when discussing the legitimacy of MEV.

For example, after a flash loan attacker manipulates the relative prices of AMM assets like Curve, arbitrageurs quickly correct the price difference, earning a substantial arbitrage profit, which is a legitimate gain for their rapid market response; whereas a sandwich attack encountered during their arbitrage process or directly copying their arbitrage trades to execute first is a form of opportunistic behavior. Similarly, bots that track market fluctuations and listen for price oracle updates to compete for liquidation opportunities fulfill the role of keepers, while listening to and copying transactions sent by other liquidation bots to act as the first liquidator robs the labor results of the liquidated transactions; some alternative trading opportunities require more effort to capture but are rendered futile in front of front-runners (the hacker involved in the DODO security vulnerability attack in March this year was also helplessly intercepted by front-running bots).

The opportunities for front-running brought about by transaction transparency prevent seekers from forming technical barriers, turning the MEV competitive market into a dark forest for front-running bots. When this highly exposed capital market attracts various front-running bots, they too must fight in the bidding mechanism for block space, going through more rounds of the PGA process to obtain prioritized transaction ordering. Mining pools, which have the power to order blocks, are naturally endowed with opportunities for rent-seeking and hold an absolute advantage in the front-running competition. This reflects an inherent flaw in the existing blockchain mechanism and pushes part of the interests of seekers towards front-runners, which are then distributed to miners through the front-runners. I prefer to refer to the total undistributed value extracted from seekers by front-runners as Maximum Extractable Value (MEV), while the additional income miners gain from the PGA is referred to as Miner Extractable Value (mev).

Negative Externalities of MEV on Blockchain Ecosystem

The existence of MEV has sparked a new competition between the seeker group and the front-runner group, providing miners with a source of income beyond miner rewards and ordinary fees. The community has debated whether MEV affects the decentralization characteristics of blockchain. One viewpoint argues that the final stage of extracting MEV is constrained by the computational power competition for packing blocks; mining pools with more computational power have a greater chance of successfully packing MEV transactions as front-runners, creating a positive feedback loop that incentivizes these pools to continuously attract miners, leading to centralization of computational power and reducing the decentralization and security of the blockchain.

However, another viewpoint counters that since mining pools cannot confirm whether they will definitely compete for the next block packing rights, they will rationally sell the MEV transaction opportunities they discover to other mining pools to lock in their additional profits, thus probabilistically distributing MEV among the mining pools participating in the MEV competition, increasing their overall profits.

From the historical trend of labor towards increasingly refined division of labor, the skills and production tools required for searching MEV transactions and mining computational power differ significantly, making the division of labor between MEV searching and computational power mining more likely. The solutions proposed by Flashbots, which will be described below, also indicate that this division of labor is forming.

A definite negative externality of MEV on the blockchain ecosystem is its malignant effect on the Gas fee bidding mechanism. As long as MEV is large enough, front-runners will be willing to pay Gas fees hundreds of times higher than ordinary transactions through malicious competition, leading to many failed high-fee transactions being recorded in block space, resulting in a waste of this scarce resource.

MEV creates a certain antagonistic effect on ordinary trading users and seekers. In traditional finance, regulatory agencies impose legal constraints on disruptive behaviors such as front-running; in the decentralized finance ecosystem, the current reliance is mainly on the game theory among stakeholders and the technological upgrades of decentralized systems to promote market mechanism reform or optimization.

Order of MEV Interest Distribution Market

In response to the negative impacts of MEV on the blockchain ecosystem, Flashbots attempts to establish an MEV interest distribution market order for seekers and miners, effectively providing a way to decentralize part of the PGA process and allowing mining pools to bid for MEV. This indeed alleviates the congestion issues caused by the MEV grabbing wars on-chain.

However, a new question arises: how to prevent various malicious behaviors from stakeholders in the off-chain system? This returns to how to ensure the privacy of transaction information before it goes on-chain; otherwise, seekers' transaction information can still be easily front-run by miners. Currently, Flashbots' MEV-Geth version employs a permissioned system to constrain participant behavior, while the latest proposal is to implement a sealed-bid auction mechanism based on SGX to achieve non-permissioned market access and transaction privacy. This is a topic worth exploring in depth, which will not be elaborated on in this article.

Figure 1. Flashbots Design Goals and MEV-SGX Architecture (from Flashbots MEV Roast 13)

Game Theory of Stakeholders

The emergence of Flashbots has changed the dark forest rules of block space, allowing seekers to choose to compete for priority ordering in an orderly market. The analysis of stakeholder game theory suggests that the new order is expected to eliminate the opportunistic chances for front-runners, enabling arbitrage and liquidation bots to engage in fair PGA competition, with the MEV distributed to miners gradually evolving into a regular price discovery premium after transaction privacy is protected. Those who employ superior strategies during the bidding process can implement the price discovery process at a relatively lower premium.

Figure 2. Two Types of MEV Acquisition Attack Methods

Currently, front-running attack transactions can be divided into two main categories:

Sandwich Attacks

Primarily target trades in AMM markets, causing users to incur greater slippage losses Cs, which is the MEV obtained by front-runners. Users can control Cs by limiting trade volume and setting maximum slippage, ensuring that 0≤MEV≤Cs. Front-runners distribute part of their MEV earnings to mining pools during the PGA.

Front-Run Attacks

Mainly target arbitrage and liquidation trades, with flash loan attacks and other trades that exploit system vulnerabilities also falling into this category. The MEV obtained by attackers in front-run scenarios equals the expected profit Re of the front-run trade, rendering the targeted trade unsuccessful or invalid, resulting not only in a loss of expected profit but also in wasted transaction fees and, in some cases, additional losses.

Back-running attacks targeting oracle updates and large trades are actually a form of clever searching behavior, causing no harm to the trades being followed; seekers merely seize profit opportunities before other followers.

In front-run attacks, the sender of the front-run trade appears to be the most innocent victim, having invested effort to discover a profitable opportunity, only to have it exploited by other competitors once made public. This phenomenon of opportunism leads to a mismatch between the labor and income in the price discovery process.

Flashbots provides stakeholders with new options.

AMM trading users and seekers are extremely passive in a model where unconfirmed transactions are transparent and public; their expected profits can be expressed as:

Re-Cgas-MEV≥0

where Re is the expected profit in the price discovery process (after deducting acceptable slippage costs), Cgas represents ordinary transaction fees, and MEV is the expected profit that may be seized by front-runners.

For front-runners, their expected profit is represented as:

MEV-mev≥0

where mev is the profit allocated to the mining pool during the PGA process.

The expected lower limit of MEV sought by front-runners is mev, while for AMM trading users, their strategy is to reduce trade volume and set slippage limits to constrain MEV.

In fact, when the Flashbots model becomes non-permissioned, users can choose to directly participate in the MEV-geth bidding process, bidding within the (mev, MEV) range, with the maximum loss being the same cost as before, or failing to bid and not executing immediately, while sandwich bots will lose many targets. If a new auction market with protected transaction privacy can be established, the issue of sandwich attacks can also be effectively alleviated.

For arbitrage and liquidation trades facing front-running, the MEV that front-runners can seize is often greater, and seekers can send these trades directly to MEV-geth nodes, protecting their transaction privacy while gaining an advantage in the bidding process (eliminating competition from front-runners and only competing with similar seekers, and failing to bid does not incur transaction fees).

For front-runners, they can still search for exploitable trades in the mempool and are motivated to send transaction bundles to MEV-geth nodes. The reason is that MEV-geth nodes can protect them from being further front-run by mining pools; in the off-chain bidding process, even failed transactions will not be packed on-chain, avoiding transaction fees or additional losses. For stakeholders who hold both front-runner and mining pool identities, as the computational power of MEV-geth nodes increases, it becomes difficult for them to maintain an advantage in opportunism. Imagine, for the same exploitable trade, an ordinary front-running bot sends it to an MEV-geth node with nearly 60% of the computational power, allowing it to finally compete with a single mining pool. As more seekers send arbitrage and liquidation trades to MEV-geth nodes in a privacy-protected manner, the opportunities for front-runners to exploit will continue to diminish.

Ultimately, the interest distribution order established by Flashbots may form an auction market dominated by arbitrage and liquidation bots, eliminating opportunistic behavior. The nature of the MEV bid by these bots transforms into a price discovery premium paid for the protection against front-running attacks.

For mining pools, there is theoretically a risk-averse preference. The official Ethereum client version has been running for years, and its security has been validated, yet nearly 60% of the computational power has switched to third-party versions, which indeed deserves attention. One reason for this may be that under the drive of EIP-1559, mining pools are gradually entering the MEV market and facing a choice—whether to participate in MEV competition themselves or adopt the MEV-geth version and follow the market order established by Flashbots. For mining pools that do not act as seekers, adopting MEV-geth can purely increase miner income and prevent miner loss; for mining pools with search capabilities, it depends on the opportunity cost sacrificed by adopting MEV-geth, for example, MEV-geth currently stipulates that only one MEV transaction can be inserted into a block, while mining pools may be able to assemble blocks containing one or more MEV transactions for greater profits. However, searching for MEV requires mining pools to possess relevant skills and consume more resources, reducing their net income.

Conclusion and Reflection

We are pleased to find that blockchain infrastructure has already paved the way for decentralized markets, and the ecosystem is beginning to iterate and develop in various niche areas, allowing us to pay attention to interesting new markets like MEV.

MEV draws our attention to the value of transaction data and the importance of block space ordering, prompting us to think about new roles beyond miners, mining pools, and users. Although Flashbots still has many areas for improvement, it provides us with a good example of how to establish order in this emerging trading market to eliminate some factors that are unfriendly to overall economic development and to build civilization in cyberspace. Under its promotion, Miner Extractable Value is expected to gradually evolve into a regular price discovery premium after transaction privacy is protected.

Just as Flashbots attempts to fork and transform the unconfirmed transaction pool module of the blockchain system, there is still room for optimization in the mechanism design of decentralized systems, which is worth further contemplation.