Utilization of MEV Blocker Kickbacks from CoW Settlements

We developed the MEV Blocker RPC endpoint in collaboration with Agnostic Relay and Beaverbuild in order to provide a very simple Pareto improvement over transaction submission via the public mempool. MEV Blocker offers frontrunning and revert protection as well as MEV kickbacks for backruns on all kinds of raw Ethereum transactions.

While MEV Blocker achieves a very good level of protection from MEV, we mainly built it as a complement to CoW Swap for transaction intents that are not yet supported by the protocol (e.g. providing liquidity, NFT trades, etc…) and as a stepping stone for integrations (giving users protection is as simple as changing your RPC endpoint).

Of course, we believe that CoW Protocol itself is still a much better approach for the DEX trading use case in particular and as a solution for MEV as a whole (wasting less block space, as well as giving more value back to users, better prices from batching and a better user experience via signed intents). In a sense, CoW Protocol is the dApp layer solution to MEV, whereas MEV Blocker is an infrastructure addition to the PBS “solution” to MEV.

Moreover, CoW Protocol itself can use MEV Blocker’s infrastructure to benefit CoW Protocol solvers and therefore also CoW Swap users. In the last week alone, solvers paid more than $20k for failed transactions (dune). Solvers are also still occasionally getting sandwiched. Note that these attacks don’t come at a direct cost to the user — who is guaranteed the price that was promised in the off-chain competition. Instead, it is more of a calculated cost solvers carry when deciding what slippage tolerance to set for AMM interactions. Depending on the current gas price & volatility setting, a slightly higher than necessary slippage tolerance can reduce the cost from failed transactions.

Therefore, we have changed the submission logic to default to using the MEV Blocker endpoint until solvers are enabled to implement their own solutions submission strategy.

Another aspect of MEV Blocker is its kickback mechanism for backruns. Searchers can bid to be placed directly behind a transaction to potentially equalise AMMs that have been touched disproportionally. The originator of the transactions receives 90% of the searcher’s bid as a refund ETH transfer within the same block.

Unfortunately, even CoW Protocol settlements are sometimes prone to backruns. The most common case is large trades which don’t attract enough private market maker liquidity and thus move on-chain liquidity sources to a point where external venues (such as Binance) offer an arbitrage opportunity. The ENS trade was such an example: https://twitter.com/0x94305/status/1623664463109816322

It is of course the goal of CoW Protocol to provide the best price across all different venues for users on Ethereum (a task solvers compete and are rewarded for). However it is the reality that those backrunning opportunities still exist today. We think it is clear that capturing 90% of such opportunities by using MEV Blocker is better than letting them go to validators, but this begs the question: how to utilize the refunds that arise? This is the goal of this post: to kickstart the discussion, which should likely result in a CIP vote. For now, the refunds are collected in this Safe which is controlled by CoW DAO. An overview of backrun transactions can be seen here: https://dune.com/queries/2403907

While we believe this value ultimately belongs to CoW Protocol users, it is hard to properly attribute and distribute the value in an automated way for multi-order batches. We also think that it would be wrong to send those refunds to the affected solvers, as they should have an incentive to improve their matching strategies in protocol and ensure users get the best price.

In the meantime, the funds could potentially be used as a revenue stream for CoW DAO to further foster the solver competition and incentivize more competitive solutions (e.g. by buying back CoW tokens to increase the economic incentive of the competition).

Please comment with suggestions and arguments on how you think these kickbacks should be used!


buying back CoW tokens to increase the economic incentive of the competition

As a COW token holder, I obviously like this idea.

What I found interesting here is that, AFAIU, the CIP-20 performance reward is paid in COW with an up-to-date price at the time of the payout. This means that, in theory, by increasing the price of COW through these buybacks, the absolve value in COW of the performance rewards would go down from payout to payout (note that the value of the payout in ETH stays the same), leaving more COW to be distributed for the consistency reward at the end of the payout period.

Not sure if this is a meaningful observation or not (aside from the fact that solvers would also benefit indirectly from the DAO using kickbacks for COW buybacks, albeit I expect significantly less than if they received the kickback directly).

token buyback is a bad idea. People benefiting from that, if any, are not economically aligned with the system to improve.
what cs needs: more traders, and more liquidity (counter trader)

  1. raise awareness (promotions, trading competitions, quizes, afilialtes) to get more trades
  2. incentives for people to put up liquidity in private pools which can be traded against

Dear Felix,

First and foremost, I would like to extend my congratulations on the successful integration. This redundancy has indeed proven to be beneficial for the core protocol participants. I have reviewed your proposal and would like to offer some insights and suggestions.

Upon analysis, I believe that the token buyback might not be the most effective approach. Instead, it would be more prudent for the DAO to retain the earned rebates and allocate them towards future developmental endeavours.

At the current stage, the protocol may not be capable of providing revenue to stakeholders. Consequently, I contend that buying back COW tokens could potentially be an inefficient use of funds for the following reasons:

  1. The COW token has limited liquidity.
  2. Past examples, such as MKR, demonstrate that buybacks may not yield the desired outcomes and could result in a misallocation of resources.

From a foundational perspective, it is crucial to consider the following points:

  1. Rebates are generated due to inefficient execution.
  2. Incorporating logic into the protocol that relies on Solvers’ inefficient execution to direct earned incentives may not be the optimal strategy.
  3. Granting Solvers these incentives could lead to a counterproductive situation, as they may then be motivated to perform inefficient swaps (assuming scenarios with minimal solver participation).
  4. While users are the rightful beneficiaries of the rebates, as you aptly noted, distribution remains a challenge.

To address the rebate utilization issue, I propose the following course of action:

  1. Establish a gas pot to be accessed by Solvers for executing their swaps. Earned rebates are then transferred to the gas pot contract.
  2. Design a well-structured logic that ensures Solvers with efficient execution receive the earned rebates, thus creating a system that penalizes poor performance while rewarding excellence in execution.

I hope you find these suggestions valuable and worthy of consideration. I look forward to further discussions on this topic.

Best regards,

At this point (after 1 week), it looks like the amount of revenue from kickbacks is luckily quite small (the Safe accrued ~$160). This is good, because it means that solvers are by and large not leaving significant back-running opportunities behind and functioning as intended (users are receiving the best possible price on CoW Swap).

However, I presume, that the kickback-distribution may end up being long tailed (ie. few high value kickbacks on very volatile days). Also, since the points brought up so far touch on more general strategic aspects wrt the protocol’s future, I believe it is good to continue this discussion.

token buyback is a bad idea. People benefiting from that, if any, are not economically aligned with the system to improve.

One reason why token buybacks in general may be interesting, is that the CoW token price is quite depressed (arguably due to the lack of utility). At this point, the protocol has sufficient treasury for a bit more than ~1y of development.
While general protocol revenue could simply be used to fund the project going forward, given that the cash flows are still small and will likely grow with increasing adoption, simply raising awareness that the CoW token may serve as a value capture instrument may put the protocol in a better position to raise funds in the future.

I agree with the moral hazard wrt the source of this particular revenue source though (in the long run we want it to be 0).

Design a well-structured logic that ensures Solvers with efficient execution receive the earned rebates, thus creating a system that penalizes poor performance while rewarding excellence in execution.

Do you have a concrete logic in mind? With the revamp of the solver rewards (CIP 20), we introduced a “participation based” reward (a fraction of rewards will be distributed based on how often solvers provided valid solution candidates to the competition).
In general, we are still struggling to incentivise solvers to build more forward looking support in their matching strategies (e.g. to this point only two solvers actively match for CoWs) and would appreciate ideas on how to improve the situation around this.