Apply Here

Problem Statement

MEV auctions leave a lot to be desired in terms of their theoretical properties and practical implementations. This RFP aims to focus on theoretical properties that aid practical implementations. A successful proposal will be able to answer a plurality of the following set of questions:

• What are the sets of properties of an MEV auction that allow for easier implementations?
• What extra assumptions do these properties impose upon the network?
• What are the security properties of this MEV auction?
  • Economic security properties
  • Game Theoretic security properties
  • Cryptographic security properties

In particular, a successful proposal will be able to tie answers to these questions to practical implementation details for an MEV auction. In the section Places to Start, we provide some explicit examples of theoretical results that answer some of the above questions that lead to practically relevant algorithms and software architectures. ******

Background

The current landscape of MEV auction design is closer to a bowl of spaghetti than a phylogeny. MEV auctions have a variety of properties that they offer to users with each auction type providing features such as:

• Centralization/Decentralization
• Permissionless bidding
• Permissionless blockspace sales
• (Partial) Privacy for bidders
• Privacy for the auctioneer
• Relay mechanisms
• Auction aggregation
  • e.g. Order flow auctions

Most MEV auctions have emerged due to the necessity rather than intentional design —  Flashbots as a response to transaction spam, Jito as a response to validator resource spam, Skip as a coordination mechanism. This means that most of the auctions have predominantly addressed practical concerns due to pressure to launch products as fast as possible. These practical measures inevitably introduce extra assumptions within these systems, such as centralized auctions or relayers or expected latency from validators or MEV searchers. On the other hand, the auction theory literature has focused on theoretical properties of auctions such as incentive compatibility. Such results allow for one to guarantee much stronger properties about auctions that ensure they maximize welfare for all participants. However, these results are extremely divorced from practice and often involve computations that are too impractical to perform in a production setting as a validator.

On the other hand, now that billions of dollars of value have been aggregated and redistributed via MEV auctions, it is becoming increasingly clear that these two disjoint worlds need to unite. Some of the live auctions have run into issues with centralization, unfairness, and/or harm to unsophisticated users. In particular, as the value transacted across blockchains has grown, so has the divide between the unsophisticated user and the sophisticated MEV searcher. The only realistic way to close this gap is to design mechanisms that utilize cryptography and/or economic incentives to ensure that the unsophisticated user preferences are satisfied in as close to an incentive compatible manner as possible.

There are a number of directions that various groups are aiming to tackle this problem. Within Ethereum and the EVM ecosystem, Flashbots’ SUAVE aims to provide a means for users to express preferences that MEV searchers have to agree upon. On the other hand, Anoma and Skip have been focusing on intents, which represent programmatic means by which users can express high level transaction constraints that searchers have to obey. However, all of these mechanisms rely in some manner on auctions that provide censorship resistance and incentive compatibility guarantees.

This RFP aims to answer the following question:

• How can we guarantee censorship resistance and incentive compatibility for auctions so that they can be core primitives of user preference enforcement via SUAVE or intents?

Answers to this question will need to address the following desiderata:

1. What is the correct definition of incentive compatibility (IC) for MEV auctions?
   1. Classical auction theory assumes that the auction knows something about the bidders’ valuation models (e.g. a distribution of values to draw from, bounds on the values being drawn, an ability to learn from multiple auctions) — but we likely cannot assume that in the decentralized, permissionless context. For instance, a bidder can change their identity in every auction making learning behavior significantly more difficult.
   2. Another question is what constraints one should use within these auctions. For instance, a common IC constraint will be something of the form

   $s\text{ is an incentive compatible strategy if }u(s'i, s{-i}) \leq u(s)\;\forall s'_i \neq s_i$

   where $s$ is a strategy profile, $s_i$ represents the $i$th participants strategy, and $u$ is a social welfare function. What sets of strategies should we be looking through for MEV auctions? What is the correct set of social welfare functions to look at?
   3. Finally, one necessarily has to consider the computability of IC conditions. If verifying that an auction is IC is a PPP-complete problem or even a hard polynomial time problem, it is very unlikely that one can check if such guarantees hold in a practical setting. How do we ensure that such guarantees hold and can be easily verified in practice?

2. What is the correct definition of censorship resistance (CR) for MEV auctions?
   1. MEV auctions could be run themselves on blockchains
      1. Credible, Optimal Auctions on Blockchains provides an example of how particular cryptographic commitments can be used to increase the set of valuations used when running an auction on a blockchain
      2. Censorship Resistance in On-Chain Auctions provides evidence of how censorship resistance can be hard to achieve if the auctioneer can collude with validators to cause a fork
   2. MEV auctions could be run within a trusted environment such as a trusted enclave
      1. SUAVE
3. How do we compare different incentive compatibility and censorship resistance definitions across different auctions?
   1. This is so that mechanism designers and chain developers can take the set of options and apply it to their context

Places to Start

• Books
  • 20 Lectures on Algorithmic Game Theory
  • Algorithmic Game Theory
• Papers
  • Credible, Optimal Auctions on Blockchains
  • Censorship Resistance in On-Chain Auctions
  • Unity is strength: A formalization of cross-domain maximal extractable value
  • SoK: MEV Countermeasures: Theory and Practice
  • Time is Money: Strategic Timing Games in Proof-of-Stake Protocols