Background
DeFi trading has evolved from vanilla onchain orderbooks to the invention and proliferation of automated market makers (AMMs), and finally to today’s off-chain auctions for transaction execution. This has led to the emergence of important marketplace dynamics that have been studied over the past few years, such as the incentives of liquidity providers in AMMs and solvers in off-chain markets.
A few key questions remain. On the empirical side: how can we compare the quality that each of these systems realizes? On the theoretical side: what kinds of guarantees can these systems (especially the off-chain ones) hope to provide to end users, and to what degree are they verifiable?
These questions are challenging to answer for several reasons:
- Simply using empirical data to compare historical realized prices across platforms may not work because it is hard to:
- sample enough orders across platforms to match the true distribution of trade preferences (or, “intent”) inputs - pairs, size, gas price
- sample orders at the same liquidity and market conditions (i.e. at the time when the trade was requested)
- Using an ideal block state simulation to benchmark may be inaccurate since in real execution environments, users’ orders compete against & collide with each other, with the presence of MEV bots activities. This leads to congestion, which is a phenomenon whose effects have been studied in earlier papers. Modeling congestion costs is therefore essential.
- Understanding the internal cost structure of solvers is challenging. The solver market has exhibited severe concentration in platforms such as UniswapX over the past year.
- Off-chain auctions are difficult to verify. In comparison, AMMs provide full visibility into their state updates.
In this RFP, we want to encourage the study of these questions from both a modeling and empirical perspective. The importance of off-chain execution marketplaces is only growing, and understanding how they interact with their on-chain counterparts is essential.
Problem Statement
Track 1: RFQ pricing quality & behaviors across different platforms
How prominent is RFQ liquidity among DeFi trading? How does its volume scale compared to AMM liquidity? How does its pricing compare to AMM liquidity across different trading venues? We would like to understand the attributes of these 2 types of liquidity across DeFi, in terms of:
- Trading pairs,
- Trading size,
- Auction platform designs
Further, how sensitive are these two types of systems with respect to market movements? Does RFQ liquidity react more sensitively towards CEX market orderbooks, while AMM liquidity pricing may remain stale with a delay until after arbitrage bots perform their orders?
How does the process of DeFi market making looking like? Do market makers take DEX orders first, and then trade on CEX; or vice versa? What are risks and inclusion guaranteee for different RFQ implementations? Understand the implementation details of each RFQ system:
- Direct RFQ Quoting: 1inch RFQ, 0x RFQ, Paraswap RFQ, Metamask RFQ integration, Bebop RFQ
- RFQ Aggregator Quoting: Hashflow
- Solver RFQ Quoting: CowSwap Solver system, Uniswap X, 1inch Fusion, Bebop Solver system
For each system, how are the RFQ soft/firm quotes get confirmed? Understanding the Retail User <> Market Maker quote flow is also important for understanding its implication towards DeFi market makers’s quoting behavior — e.g. quoted price spread varies depending on:
1) If the Market Maker has the first-look or last-look, i.e. submission/cancellation control?
2) How long of expiry does the signed RFQ order have — is it the longer the worse the pricing is?
3) How does the vertical integration in block building infra layer affect its pricing? (This is a sub-topic in connection to the Block Bidding RFP of this same cohort.)
Track 2: Execution quality and verifiability across DEX trading platforms
Multiple factors can affect a DEX platform’s execution quality: gas efficiency of routers, liquidity sources, pricing comparison algorithm (or atlas the auction design), gas estimates and default slippage threshold from UI. A few studies have looked into users’ slippages across major DEXes and aggregators, which revealed the presence of MEV activities as well as the power of default settings. But there hasn’t been any conclusions in consensus about “Which platform has better execution outcome, combining all factors?” Is the off-chain solver model really more superior than aggregators, who then theoretically provides better pricing than naive DEX routing?. Empirical study or simulation benchmarks can bring great impact and clarity for the whole industry to understand current state and “if we are really progressing DEX routing design” — such execution quality needs to be validated in data!
Particularly, Solver (or Intent) Models (Uniswap X, Cowswap, 1inch Fusion) as the latest novel design draws lots of interests. How do current designs and their performances compare to each other?
- How does each implementation differ in terms of permission-ness, submission control, settlement flow (e.g. OFA usage like MEVBlocker, builder preference and integration)?
- How does each design favor different liquidity?
- How does the scale and pricing compare across each other?
- How are liquidity vertically integrated across platforms?
- In a world of a unified entrypoint posting orderflow across — how much price improvement can happen from internalizing users’ orders? (ie. network effect from Coincidence-Of-Wants)
Places to Start
- Papers
- Data Sources
- orderflow.art
- Videos
