Illuminating the Dark Forest: Unveiling the Mysteries of MEV

With the surge of activities on the Ethereum chain and the evolution of its infrastructure, MEV has been seen as one of the most dangerous aspects of the Ethereum ecosystem, directly causing profit losses and degraded experiences for users. This article aims to analyze the centralization and trust issues brought about by this mechanism and the contradictions with Ethereum's decentralized values, starting from the block generation mechanism of Ethereum 2.0 and the technical evolution of proposer-builder separation (PBS).

MEV is indeed a double-edged sword, with both positive and negative externalities. The positive aspects include reducing price discrepancies on DEXs and assisting in transaction liquidation; the negative aspects include harm to users from sandwich trading. Therefore, solutions for MEV are more about mitigating negative externalities rather than completely eradicating them. In exploring how to alleviate the negative externalities of MEV and address the current issues with trust-based third-party middleware Relayers, measures can be primarily categorized into three types: auction mechanism improvements, consensus layer improvements, and application layer improvements. These improvements will affect the MEV landscape to varying degrees, but some proposals may not fundamentally resolve the sandwich attack problems that users face. User transactions remain in a public pool, thus more privacy pool technologies need to be introduced to protect the optional privacy of transactions, and these MEV solutions are worth trying in combination.

In addition, MEV, as an unavoidable byproduct of mechanism design, will become even more complex in the future. We also discussed the potential MEV technical challenges and opportunities that may arise under the implementation of new transaction types such as Layer 2 architecture and EIP-4337 account abstraction.

Finally, this article hopes to explore potential solutions to mitigate the negative externalities of MEV, and to provide a comprehensive understanding of the pros and cons of current MEV solutions, illuminating the dark forest in which users find themselves for the future, as well as guiding industry researchers towards further studies on MEV.

Ethereum 2.0

Since The Merge, Ethereum has adopted the POS mechanism to ensure network security, abandoning computationally intensive competition for block production in favor of proof of stake. After the merge, Ethereum is divided into an execution layer and a consensus layer. Block production has also changed, with each Epoch representing a POS cycle, each Epoch divided into 32 Slots, and each Slot being a 12-second block time unit.

Each Epoch of the network randomly selects a committee, from which a block proposer is randomly chosen to package and sort transactions and produce blocks, while other committee members supervise and vote. The committee is re-selected after each Epoch, with a time limit on operations to ensure efficiency. Here, the term Payload is standardized as the execution load, which refers to the state changes of transactions and can be seen as part of executing a block. The block proposer will implement the execution load and the block proposal.

PBS Architecture

In fact, when validators are selected as block proposers, there is often little incentive to execute the Payload, because it requires a significant amount of computational power. The original idea was to decentralize transaction ordering through elections by a decentralized committee, but validators tend to outsource this, focusing on proposing blocks themselves. This led to the concept of PBS, which separates block proposing and construction, where proposers are only responsible for validating blocks and do not participate in construction. This promotes an open market where proposers can obtain blocks from builders. Builders compete with each other to construct blocks, offering the highest fees to proposers, known as "block auctions."

PBS(Proposer Builder Separate)The sealed first auction model process is as follows: Users submit transactions to the public Mempool through RPC proxies, and multiple Builders find suitable transaction orders to generate profit-maximized blocks(including Base + Priority + MEV), which are then interacted with the Proposer via MEV-Boost Relayer. The Relayer acts as a bridge, Builders submit bids to it, and the Relayer submits multiple block headers and bids to the Proposer, who usually adopts the highest bid. The Relayer implements the MEVBoost specification, regulating the bidding interaction between Builders and Proposers. During this process, information is sealed, and the Relayer only submits block headers to the Proposer, ensuring censorship resistance.

Participants and Game Theory under PBS

The main participants are Builder, Relayer, Proposer, MEVbot(Searcher).

Builder

Builders are responsible for constructing block content, and using MEV-Boost is more advantageous in bidding because it supports MEV profits. Builders can directly review transactions, which has been criticized, especially after the announcement by the US OFAC, leading to a large number of Builders participating in OFAC Compliant activities. Although the recent proportion of reviewed blocks has decreased, Builders still play a direct role in transaction review.

Currently, the share of the Builder market is gradually expanding for beaverbuild.org, which requires no review, all oriented towards profit.

Searcher

Maximizing profits requires the joint efforts of Searchers and Builders. Searchers often collaborate with specific Builders to form Dark Pools or Private Pools, where Searcher transactions are only visible to certain Builders. Some Builders obtain MEV transactions that maximize profits to bid for block space. Theoretically, if a Builder acts maliciously or censors, the Searcher can choose other Builders, leading to a decrease in market share for the malicious Builder. Therefore, Builders will consider the implicit costs of acting maliciously.

Searcher is divided into two main categories: CEX-DEX( off-chain) arbitrage and pure on-chain( DEX, mezzanine, and liquidation). Currently, Wintermute holds the largest market share in CEX-DEX arbitrage trading. Pure on-chain MEV opportunities are trending towards studio-style, with jaredfromsubway.eth capturing 37.2% market share, excelling in sandwich attacks, and once becoming the highest gas consumer on-chain. Due to the close relationship between Searchers and Builders, many Searchers route order flow to the top three Builders to maintain ecological influence and avoid the risk of strategy failure caused by small Builders splitting order flow.

Relayer

The Relayer is responsible for aggregating bids and acts as an intermediary to submit the block header and bidding price to the Proposer, who is unaware of the transaction details at this point. After the Proposer selects and signs the block header, the Relayer releases all transaction content. The Relayer, as a third party without economic incentives, gains significant trust, and historically there have been vulnerabilities that led to the Proposer extracting over $20 million in MEV. While the vulnerabilities can be patched, the Relayer may still act maliciously to steal MEV.

Currently, the market share of builders running pure MAX Profit has gradually expanded since the Merge, and in the free market, MEV cannot be artificially controlled by builders. Relayers also face the issue of lack of economic incentives, and Blocknative has exited Relayer development. Relayers are built based on the MEVBoost specification proposed by Flashbots, and relying on third parties to provide PBS for Ethereum is not a long-term solution; the community is exploring the inclusion of PBS at the protocol level.

Proposer

The Proposer randomly selects from all validators, possessing execution load capacity but inclined to outsource, which can easily lead to vertical collaboration with the Builder. The Relayer of MEV-boost hopes to serve as an intermediary point to reduce the vertical collusion caused by direct communication. Currently, mining pools act as validator pools, and the emergence of LSD enhances capital efficiency, leading to a trend of centralization in validator pools.

Lido currently holds 28.7% market share, with Coinbase and Ether.fi ranking second and third. In the past, when MEV-BOOST PBS was not implemented, the Proposer was responsible for the Builder tasks, but most gave up their transaction ordering execution capability due to the heavy computational load affecting validation performance, opting instead to outsource the execution load to third parties for block auctioning.

User

Users are in the weakest position throughout the entire architecture, with transactions placed in the Mempool for MEV bots to profit, but these profits do not flow to the users. However, in DEX, MEV bots can also mitigate slippage and platform price differences for large trades through arbitrage. Therefore, MEV has both positive and negative externalities, which need to be discussed separately.

To prevent users from being harmed by MEVbots, many RPC node providers help users place transactions in a private Mempool, such as interacting directly with Builders. Novel methods like OFA(Order Flow Auction) compensate users for MEV profits through order flow auctions, collaborating with Searchers to maximize MEV and return it to users.

Currently, the proportion of users using private order flow is about 10%, mainly due to high education costs and complex operations. Optimizing user experience requires users to passively accept more rather than actively engage.

Summary

Under the current PBS architecture, since the introduction of the MEV-BOOST specification, the profit-maximizing sealed-bid auction mechanism has led to a gradual cooperation and trust between Builders and Searchers, with a clear trend towards centralization. Validator centralization under POS has made all links in the MEV industry chain centralized, introducing multi-party trust issues. The centralization and trust-based development of MEV contradicts the decentralized and trustless vision of Ethereum. The Ethereum community is currently discussing three proposals to mitigate centralization:

1. Regarding the bundling of Builder and Searcher centralization: Flashbot proposed the SUAVE technology to increase transaction transparency, lower the trust threshold of Searchers towards Builders, and encourage Searchers to send order flow to all Builders.

2. For Relayer trust: use Enshrined PBS instead of the current PBS scheme to eliminate reliance on Relayer.

3. Regarding Validator centralization: adopting decentralized AVS, such as SSV, Lido has already partnered with it.

Current Status of MEV

Currently, the main MEV on-chain includes arbitrage, sandwich attacks, and liquidation. In the past 30 days, arbitrage profits have been the highest, with MEV bots reportedly making a profit of 2.6 million USD. The average profit per transaction is 0.8 USD, and the profits from sandwich attacks on the Ethereum chain in the past 30 days are approximately 880,000 USD.

The positive external effects of MEV include reducing price differences between DEXs through arbitrage and helping DeFi protocols liquidate collateral, among others. The negative effects mainly harm user profits due to sandwich trading. Under the current on-chain fee mechanism, although Ethereum has implemented Gas Fees smoothing, the increase in on-chain arbitrage opportunities still leads to short-term spikes in Gas Fees when MEV bots trade alongside users, resulting in economic losses and a poor user experience.

In addition, the transition of Ethereum to the Layer 2 architecture has also given rise to cross-chain MEV issues between Layer 2s.

Potential MEV Complexity in Layer 2 Architecture Design

Future large-scale on-chain arbitrage activities will shift towards more complex and technically demanding multi-chain cross-chain MEV. Currently, there is limited research, but some countermeasures have already been proposed, primarily focusing on improvements to the orderer. Cross-chain bridges are essential products for cross-chain operations between different Layer 2s. Searchers can help mitigate the fragmented liquidity between Layer 2s, but the current impact is not significant, mainly due to the need for improvements in cross-chain bridge experiences and security, as well as the higher threshold for strategies due to different finalities of various bridges.

The potential MEV of EIP-4337

EIP-4337 introduces account abstraction and new transaction types, which will also lead to significant changes in the MEV landscape. The newly added User Operation transaction type in ERC-4337 is packaged into a normal transaction by Bundler after entering the Mempool. Once the user's transaction enters the public pool, it will be monitored by Searchers. Bundler is similar to Builder and can collaborate with Searchers to rearrange User Operations to capture MEV. The specifications for User Operations may vary across different chains, further raising the technical threshold for cross-chain MEV.

MEV Mitigation Exploration Direction

In the past, within the Ethereum ecosystem, the PBS solution was outsourced to Flashbots, with its latest valuation reaching $1 billion. However, Relayers have no economic benefits and a high implementation threshold, leading Blocknative to abandon this track. To address the issues of trustlessness and zero economic incentives, Ethereum is considering using e-PBS protocol-level improvements to avoid relying on third-party protocols like mevboost for Relayers.

Currently, MEV seems difficult to completely resolve, as it is an inevitable product of the increasing complexity of the ecosystem and information asymmetry. Ethereum cannot completely eliminate MEV through protocol-level review and improvement. More efforts are focused on alleviating the negative externalities of MEV and enhancing the positive externalities. Many projects, community members, developers, and VCs are exploring potential solutions. Next,