Simple Summary
This CIP removes three social consensus rules which have been superseded by additions to the mechanism of CoW Protocol. The rules which will be removed are the rule prohibiting overbidding, the rule requiring Ethereum Best Bid and Offer (EBBO), and the rule requiring local token conservation. Other social consensus rules are reformulated to better capture the intent of the rule.
Motivation
One recurring criticism of the rule set of CoW Protocol is that it is more complicated than it needs to be, making it difficult for solvers to join the competition. This CIP removes some social consensus rules which complicate the mechanism without providing proportionate benefits.
The rules proposed for removal have been around for a long time and have in the meantime lost their value.
-
The rule on overbidding was introduced in CIP-13 to prevent pennying when the reward for winning an auction was essentially fixed. Current rewards based on second-price auctions make accurate bidding a more profitable strategy in most cases.
-
The rule on prohibiting the provision of unfair solutions, sometimes referred to as EBBO (Ethereum Best Bid and Offer) was introduced in CIP-11 to prevent surplus shifting in batched solutions. It was later formalized and adapted to all executions which did not satisfy a certain standard. Fairness filtering for combinatorial auctions prevents surplus shifting. In combination with multiple winners, fair combinatorial auctions provide a much more general and powerful baseline for executions.
-
The rule on local token conservation was also introduced in CIP-11 to prevent a form of surplus shifting observed at the time. Again, surplus shifting is more generally addressed using fair combinatorial auctions.
Specification
Rational
While social consensus rules cannot easily be removed completely, they should be limited to a minimum. They are typically hard to enforce, create risk and the opportunity for gaming from solvers, and maintenance overhead and the opportunity for executive overreach from the DAO and core team. For each rule, these disadvantages must be weighted against the protection they bring to users, solvers, and the DAO. Whenever possible, social consensus rules should be replaced by hard rules enforced in the competition.
The rules proposed for removal are of a form covered to a substantial amount by hard rules of the competition.
-
Pennying/overbidding (CIP-13):
Overbidding is a difficult concept to define. The rule can only be checked in cases where an honest routing is known. This is only (partially) true for solvers using on-chain liquidity. Most observable overbidding revolves around not taking execution costs into account. Preventing this requires deep knowledge of fair prices which the protocol cannot reliably provide. Overbidding can therefore only reasonably be enforced for solvers using on-chain liquidity. Overbidding by private market makers cannot be reasonably detected. The limited benefit of this rule thus leads to downsides for solvers which are particularly useful to the ecosystem.
At the time of CIP-13, the reward for winning the solver competition was essentially fixed, more precisely affine linear in the number of settled orders. If all solvers are similarly strong, a solver could then overbid to win the competition and be rewarded that full amount. This was feared to lead to a situation where solvers do not win by providing better routing but only due to artificial inflation of the score in the competition.
The current reward mechanism of CoW Protocol is designed to make truthful bidding a profitable strategy.
There are still cases where bidding more or bidding less than what a solver can provide in a cost covering way is even more profitable. For example, consistency rewards can lead to it being profitable to settle arbitrage orders which might otherwise not get executed. A rule prohibiting overbidding can lead to a healthier solver competition since it prevents pushing all rewards to users.
Since any benefit of overbidding goes towards users removing this rule is not an immediate threat to and potential downsides will only materialize through long term effects on the solver competition. As soon as a solver intentionally behaves in a way which fundamentally hurts the protocol, the catch-all rule on malicious behavior can be used as well. -
Provision of unfair solutions (CIP-11, CIP-52):
This rule was originally put into place as a solution to surplus shifting. It was meant to enforce fair exchange rate in CoWs where the exchange rate is known and the rate on CoW Protocol was not uniquely determined by maximizing surplus. It was later formalized and adapted to prevent all types of executions below a minimal standard (see CIP-52).
Fairness filtering in combinatorial auctions prevents unfair executions as long as there is one solver providing this minimum. This filtering prevents surplus shifting more generally as long as some solver is able to provide such a baseline. If no solver is able to provide this minimum reliably, which can happen in case of new pools or esoteric splitting across lots of pools, defining EBBO as a minimal standard seems unrealistic. Solvers suffer unbounded risk from executing orders, potentially preventing participation and making the competition less robust. The current rule does not include all liquidity sources which are required to provide good prices in the current strong competition. If a similar rule adapted to the current liquidity landscape would be required, it would be better to invest resources to build a solver following this rule, immediately setting this minimum as baseline for all other solvers.
Predatory solvers always trying to settle at unfavorable prices and depending on other solvers not participating in an auction can already be sanctioned based on the more generic rule on other malicious behavior. Only solvers with malicious intent would be penalized then, which is in line with the original goal of this rule. -
Local token conservation (CIP-11):
The original purpose was to contribute to solving the surplus shifting problem. Even though it is not applicable to all auctions due to issues in the mathematical formulation, it can prevent a certain type of surplus shifting. It does not catch the prototypical surplus shifting in a direct CoW between two sell orders, though. Fairness filtering in combinatorial auctions provides a better guarantee for that common case. The additional protection from surplus shifting can hardly outweigh the significant complexity introduced by enforcing this rule on the solvers’ and protocols’ side.
This proposal intentionally keeps some social consensus rules in a potentially reformulated form.
-
Inflation of the score (CIP-11)
The reason for this rule staying relevant for now is the problem of solvers intentionally creating order flow or otherwise influencing the auction to reap a reward on it. While this is also difficult to check, we have not found a clean solution to preventing this on the mechanism level. For now, the rule is reformulated to not single out the creation of tokens over other means of manipulating rewards. -
Illegal use of internal buffers (CIP-11)
The protocol still benefits from giving solvers access to internal buffers, e.g. for covering slippage from AMMs or to save on gas costs for small trades. At the same time, buffers pose a substantial inventory risk and solvers need to take full responsibility for access to buffers through them.
Since the DAO should not act as a formal liquidity provider by defining precisely conditions and prices for usage of buffers.
The rule is therefore kept and reformulated to better capture the intent of the use and the restrictions of use of buffers. In particular, the core team reserves the right to use any reasonable price vector to convert buffer values in slippage accounting. -
Other malicious behavior (CIP-11)
This rule is a catch-all clause for malicious behavior. While being very brought in scope, it is limited by the fact that full executive power lies with the DAO.