## Grant title Surplus-capturing AMMs --- a feasibility study ## About the team **Andrew** : Ph.D. Mathematics, 10+ years experience academic research at internationally leading pure research institutions. [MEV research](https://github.com/flashbots/mev-research/blob/main/FRPs/active/FRP-22.md) w/ Flashbots. [awmacpherson.com](https://awmacpherson.com/) **Evan**: Mathematician, Data Analyst, DeFi Mechanism Design researcher. Previous research includes bonding curve controller modeling, analysis of RMM-01 LP mechanics, analysis of MEV arbitrage mechanics on POL, and cross margin risk modeling. [evandekim.eth](https://mirror.xyz/evandekim.eth/kowg_VFD7lp5p12C4wcytc2rooVXgKnUwBd-KUKtndQ) **Mohammed**: Electrical/Electronics Engineer. MSc Economics focus on Asset pricing and credit cycle modeling. 5+ years in TradFi derivatives trading. Currently researching and building digital asset focused algorithmic trading/market making algorithms, PBS research and post-merge ETH testnet development (Ephemery/Holesovice). ## Grant category Analytics and Integrations ## Grant description ### Background The proposal builds on a [thread in the COW forum](https://forum.cow.fi/t/cow-native-amms-aka-surplus-capturing-amms-with-single-price-clearing/1219) about a CoW-specific form of [MEV-capturing AMM](https://ethresear.ch/t/mev-capturing-amm-mcamm/13336). The discussion surfaced exciting prospects for AMM designs that promise to take back control of "surplus" value ordinarily captured by arbitrageurs and MEV searchers. This value can then be deployed, for example, to achieve better prices for users and an improved overall return profile for LPs than the traditional CFMM. ### The surplus distribution problem In the classic CFMM design, prices are updated by arbitrageurs who are paid for this service by capturing the price differential between the CFMM and external markets. Liquidity providers (LPs) effectively bear these costs of this by constantly having to accept trades at unfavourable prices, a phenomenon known to economists as adverse selection. Although these costs can be offset by trading fees, this approach to value allocation is limited in two ways: - Linear trading fees equally impact arbitrageurs and “noise traders” who may be prepared to trade at prices favourable to LPs. - Arbitrageurs compete only for positioning of their transactions in the block, and the proceeds of this competition goes to block producers rather than LPs. There is no room for a competitive market for updating prices. The CoW batch auction provides a more flexible approach to delivering price information which distinguishes arbitrageurs (or “solvers”) from noise traders and permits new dimensions of competition on price discovery. On the other hand, the CoW protocol as it stands does not provide its own liquidity, and in practice most of the liquidity comes from standard CFMMs whos LPs suffer adverse selection by trading with the CoW solver. The basic idea of this proposal is to evaluate the feasibility and impact of an approach to AMM design which custodies its own liquidity but uses the pricing discovered by CoW solvers. Such an AMM would be quite remarkable in that it would provide a way to earn yield for LPs from passive liquidity provision without exposure to adverse selection at the some time offer better pricing for CoW users. Given the power of this result, it is natural to expect that some challenges lie in its realisation. In our view, a feasibility study is necessary because of apparent confusion about the nature of the value that could be captured by such an AMM, undefined aspects of its design, and the [security](https://samczsun.com/so-you-want-to-use-a-price-oracle/) [implications](https://jumpcrypto.com/so-you-still-want-to-use-a-price-oracle/) of using the CoW solution as a price oracle for an external protoocol. ## References - Price volatility impact on LPs studied by Milionis et al. in LVR (paper titled: Automated Market Making and Loss-Versus-Rebalancing) ([https://arxiv.org/abs/2208.06046](https://arxiv.org/abs/2208.06046)) - Presentations by authors discussing LVR further ([https://www.youtube.com/watch?v=G8qxxlqvMBk](https://www.youtube.com/watch?v=G8qxxlqvMBk)) ([https://www.youtube.com/watch?v=q5vyJJb-Uyw&list=PLEGCF-WLh2RK2Cym4ZeEWDOSlmirXWtZO&index=34](https://www.youtube.com/watch?v=q5vyJJb-Uyw&list=PLEGCF-WLh2RK2Cym4ZeEWDOSlmirXWtZO&index=34)) - Surplus capturing AMM can be traced back to research proposal on MEV-capturing AMM designs ([https://ethresear.ch/t/mev-capturing-amm-mcamm/13336/22](https://ethresear.ch/t/mev-capturing-amm-mcamm/13336/22)) ([https://ethresear.ch/t/mev-minimizing-amm-minmev-amm/13775](https://ethresear.ch/t/mev-minimizing-amm-minmev-amm/13775)) - Paper provides formal specification of implementation of MEV-capturing AMM ([https://arxiv.org/pdf/2210.10601.pdf](https://arxiv.org/pdf/2210.10601.pdf)) - Paper details math behind using external price oracle in standard CPMM ([https://anrg.usc.edu/www/papers/dynamicautomation.pdf](https://anrg.usc.edu/www/papers/dynamicautomation.pdf)) - Forum proposal continues ideas on McAMMs, using fees as main tool for capturing LVR for LPs ([https://ethresear.ch/t/dynamic-mev-capturing-amm-dmcamm/13849](https://ethresear.ch/t/dynamic-mev-capturing-amm-dmcamm/13849)) ## Grant goals and impact - Propose high-level candidate architectures for “surplus-capturing” AMMs. In particular, address whether changes to CoW protocol itself are needed. Investigate security considerations, economic and otherwise, of the proposed architectures and discuss mitigation strategies. - Develop a quantitative model of the additional value that could be captured and delivered to LPs and users and measure it through empirical studies and simulations. **Deliverable** A technical report outlining relevant theoretical background, design considerations, metrics, and results of empirical studies. GitHub repository containing the code used for data studies. **Impact** The study will provide evidence as to whether current COW tools such as the competitive solver pricing can be utilized to build a surplus-capturing AMM in practise. It will also analyze the kind of "surplus" it would capture, its size, and the extra value users and LPs expect to receive. Thus preparing the ground for detailed design implementation. ## Milestones - Architecture. Specify candidate architectures (which will be used to define simulations and economic attacks). Discuss: - How prices are communicated from solver to AMM. - How LP portfolios are managed. - Block/batch synchronisation. - Security. Study attack scenarios and propose mitigation strategies. Discuss fallbacks in case of a delayed or missing oracle update. - Quantitative. Compare portfolio metrics of LPs in proposed designs with those of a traditional CFMM using historical market data. Clarify the relation of these quantities to LVR. - Quantitative. Evaluate the use of COW solvers as an oracle for an arbitrage-free price by comparing the performance of COW solvers with other approaches to computing this price. ## Grant timeline 6--8 weeks ## Budget breakdown Upfront: 15k XDAI Upon completion: 6k XDAI / 150k COW (12month linear vesting) ## Gnosis chain Address To be disclosed to grants team
Last changed by  
 
ek247375
122

Read more

Baseline Protocol Simulations Analysis

sdf

2/2/2024
COW tokenomics

Summary COW token was launched in January 2022 as a governance token for the CoW protocol (and made tradable a few months later) with an initial supply of 1 billion tokens. It has been distributed to various ecosystem participants and is regularly paid out (from the DAO treasury) to solvers who win batches. Currently, apart from its use as a governance token, COW does not appear to have any utility. COW Protocol Timeline GIP-13. (2022-01-06). Spin off CoWDAO from Gnosis and issue non-transferable vCOW governance tokens. https://forum.gnosis.io/t/gip-13-phase-2-cowdao-and-cow-token/2735 CIP-2. (2022-03-11). DAO will pay 100 COW tokens each batch to the winning solver. https://forum.cow.fi/t/cip-2-solver-rewards/488

5/9/2023
Data Pipeline (section 5.1?)

We collected on-chain data from two Uniswap v3, CoW, and Chainlink price subgraphs. To perform this task, we utilized the DataStreams python library, which is a subgraph query utility package. We performed merges using Polars, following the Apache Arrow standards. This standard is an emerging in-memory columnar analytics that can accelerate data load times, reduce memory usage, and speed up calculations. We restricted the data to WETH/USDC orders on CoW, Univ3 USDC/WETH .05%, and .3% fee pools, and for ETH/USD Chainlink price data. To obtain the Univ3 data, we queried the swaps schema for the USDC/WETH .05% and .3% fee pools. We merged the CoW trades data and Univ3 Swaps data for transactions at the same timestamp, adding decimal places for the amounts and calculating marginal swap execution prices for comparison. Misc Notes, needs to be organized trade directionality - we label the trade direction WETH -> USDC when WETH is in the CoW buy column and the UniV3 amountOut column. Conversely, the USDC -> WETH direction is when USDC is in the CoW sell column and the Univ3 amountIn column. the gas-adjusted price - we adjusted the execution price by adding the wei cost at each block if it is a WETH buy and subtracting if it is a WETH sell since the CoW solver execution price reflects the gas cost, whereas Univ3 and Chainlink do not. query optimization - Univ3 pools have a lot more swaps so we filter based on CoW timestamps. However since Chain has a lot less frequency than CoW, we don't apply this filter or need it as an optimization step. Forward Fill - In order to construct the final price dataframe, we create an index ranging from the CoW first and last trades. Then we forward fill Univ3, Chain, and CoW execution prices to the index to propagate the last valid observations forward through time.

3/10/2023
Decentralized Options Clearing House on Qredo

Qredo Qredo combines a fast-finality blockchain (Layer 2) for digital asset tracking and settlement with a Consensus-Driven Multi-Party Computation (CD-MPC) network, and a secure, end-to-end encrypted decentralized conversation replication network (Layer 3) to handle everything from machine to machine communications to storing and providing an audit trail for regulated pre-trade communications. A distributed ledger is used to record the ownership of a Layer 1 crypto asset, which is represented by a synthetic token/wallet combination on the Qredo blockchain and maps one to one in value to the asset’s wallet on the Layer 1 blockchain. Decentralized RFQ Qredo’s Layer 2 / Layer 3 design includes an automated, decentralized RFQ bot that is built into the Version 2.0 protocol. Qredo’s decentralized RFQ (Requests for Quotes) system securely interacts with Market Makers and enables private pre-trade negotiations between a Trader User and another known counterparty to finalize a P2P transaction, enabling a ‘Bloomberg Chat’ like experience for all users. The RFQ bot enables a simple Uniswaplike Web3 interface to facilitate the exchange of Layer-1 assets (BTC / ETH) using Market Makers that have staked and bonded QRDO tokens with a Validator. Market Makers take part in the token rewards scheme and are incentivized to offer the best prices and market depth. Any trader on the Qredo network can create atomic swap transactions with any DEX that Qredo can access and a) arbitrage trade across multiple DEXs and b) completely offload the counterparty delivery and settlement risk to Qredo Loan Pools, for a fee. In other words, a trader can create atomic swaps between DEX’s on the same Layer-1 chains (ETH) or even between different Layer-1s (ETH to BSC, ETH to BTC, etc.) without being detected by the Dark Forest and have immediate settlement. Multi-leg strategies are also easily executed with this capability. The protocol charges traders 0.5 basis points of both maker’s and taker’s trading principal. If for example two traders were exchanging BTC for ETH, the protocol would obtain ETH and BTC. The protocol exchanges Layer 1 assets for QRDO tokens using the protocol’s Request For Quote mechanism which queries both Market Makers and DEXs on multiple networks.

10/11/2022
Read more from ek247375
Published on HackMD