Host: Stephanie (@stephaniiee_eth)

 Guest: Anna (@AnnaMSGeorge)

If you're interested in DeFi and MEV, you may have heard of intent-based trading, frequent batch auctions, solver models, and order flow auctions (OFAs). CoW Swap can be considered a prime example of these innovations, seamlessly integrating them into DEX products to provide users with the best prices, find optimal routing, and protect against MEV attacks.

In this episode, Stephanie and CoW Swap co-founder Anna discuss in detail the design of CoW Swap, including how the trading cycle that starts with collecting user transaction intentions differs from the traditional trading lifecycle; how to ensure the safety of user funds when outsourcing trade execution to established solvers; how to foster effective competition among solvers, which is a key component in elevating intention-based trading to a new level; and how to continuously incentivize solvers while avoiding their malicious behavior?

They also discussed broader topics, such as how newcomers perceive Uniswap's first-mover advantage, why CoW Swap adopts a MEV minimization approach instead of a MEV maximization approach, and why MEV mitigation at the dapp level makes sense.

 "CoW Swap's Trading Cycle and Its Design"

 The trading cycle of CoW Swap is mainly divided into four steps

Users express their intentions through signed messages, which include the tokens to be bought or sold, the quantity of the transaction, and the validity period for executing the trade.

2. The CoW protocol collects user intentions and places them in an off-chain order book for batch auctions.

Third-party transaction executors, known as solvers, access the order book. They excel in mathematics and can quickly identify the optimal execution path (including all on-chain liquidity, demand coincidences, and circular trades) and the best prices. They compete with each other, and the protocol ranks them based on the value they can provide to users. The winning solver in the auction can execute the batch of transactions.

The winning batch executes all intents on-chain at a uniform price within a single transaction.

How CoW Swap Initiates Competition Among Solvers

The reason CoW Swap can elevate intention trading to a new level lies in the introduction of competition at the execution layer, rather than just a simple on-chain execution model. For this competition to be effective and to ensure users' trades receive better execution, the development of CoW Swap has primarily gone through three stages:

Gnosis runs some solvers internally, starting off relatively simple by collecting API endpoints from aggregators like paraswap, 1inch, and 0x. It continuously compares their return values and then decides to submit the transaction to the API that can provide the best trading return.

During last year's Devconnect in Amsterdam, I encountered a very smart and mathematically skilled team. They were small in size and very interested in running solvers, so they began developing their own algorithms and successfully triumphed in the API competition.

Market makers are starting to take an interest in solvers; they are beginning to integrate with existing solvers by providing their own liquidity, which gives them a competitive edge.

Currently, the CoW Swap solver pool consists of 16 solvers, each with their own areas of expertise.

 The coexistence of motivation and constraints in design

CoW Swap distributes incentives to solvers every week, which are divided into two parts: one part is related to continuity, as you want solvers to continuously participate in the competition, rather than only bidding when they feel they can win in a particular round. Additionally, it ensures that there are still solvers bidding even when some are offline or acting maliciously. The other part is based on how much better the solution found by the winner is compared to the second-best option. This is to ensure that they do not focus solely on providing a path that is just one cent better than the second-best; the more value they find beyond the second-best, the greater the incentives they receive.

The source of incentives currently comes from 2% of the annual distribution of CoW tokens. In January, we will introduce a small fee, which is essentially derived from the value we provide to users. These fees are earned by solvers, but they must use them to repurchase CoW tokens and return them to the CoW treasury.

Currently, the solver set theoretically does not require a license, but in practice, there are some elements that do require one. Since solvers can theoretically utilize the user's slippage tolerance, CoW Swap requires participating solvers to provide collateral. This mechanism does not have to be centralized; the current centralization is due to CoW DAO managing the collateral pool to lower the entry barrier. However, in 2024, solvers will be able to create their own collateral pools, and smart contracts will automatically check whether the solver has a private key, whether access permissions are provided, and whether a collateral pool has been established. This way, solvers can automatically join the competition.

"MEV Maximization vs. MEV Minimization"

CoW Swap focuses on minimizing MEV, as methods for maximizing MEV carry certain risks.

First, from the very beginning of maximizing MEV, you need to extract value from users. To extract value, you need multiple parties involved, and then you need to reward their efforts, meaning users cannot reclaim 100% of the MEV. In the best-case scenario, users can receive a small rebate, but from CoW Swap's perspective, the value of MEV is created by the users from the start, and they should not incur any losses.

Secondly, due to the need to reallocate value, it will also become inefficient, as it requires bundling the refund transactions into the blocks, which means more block space is needed.

Finally, MEV maximization has brought more complexity, and now there is a need to introduce PBS (Proposer/Builder Separation). Flashbots initially brought MEV to the forefront of discussions and advocated for its democratization, meaning that everyone should be able to extract MEV rather than just a select few players benefiting from it, which is a positive development. However, the downside is that it has allowed more players to enter the game, making MEV extraction truly professionalized, which has exacerbated the issue of MEV today. Of course, we also don't know what would have happened without these discussions, but it has indeed become a significant problem now.

CoW Swap believes that most MEV opportunities occur at the application layer, where users initiate transactions, and therefore, efforts should be made to minimize the chances of MEV occurring. CoW Swap collects user transaction intentions and processes all transactions in a block at a uniform settlement price, ensuring that there is no possibility of value extraction or reordering of these transactions.