CIP-36: Adjusting and renewing solver rewards budget

CIP: <to be assigned when moved to phase 2>
title: Adjusting and renewing solver rewards budget
author: Haris Angelidakis, Andrea Canidio, Felix Henneke
status: Draft
created: 2024-01-19

Simple Summary

The previous solver rewards budget was defined and approved in CIP-29, about five months ago, committing 10M COW tokens for the solver competition, projected to last until the payout of January 30, 2024. The estimate of CIP-29 indeed proved to be accurate and the allocated tokens will be depleted after the payout of January 30, 2024. This CIP is intended to propose an adjustment of the caps specified in CIP-20 and of the per-quote reward specified in CIP-27, and, moreover, propose a renewal of the solver rewards program by allocating an additional 8M COW tokens, projected to last for 6 months; specifically, from Tuesday, February 6, 2024 until Tuesday, July 30, 2024.


There are 18 running solvers at the time of writing, consistently competing to provide the best possible solutions to the batch auctions the protocol is running every few seconds, and consequently ensuring best execution for CoW Protocol users. Moreover, for the past 6 months, and after CIP-27 was introduced, multiple solvers have also become so-called “price estimators” for the protocol, and compete to provide the best possible quotes, accessible via the API to multiple integrators (e.g., CoW Swap, DefiLlama), so as to attract as much orderflow as possible. Overall, the solver competition over the past several months has seen great advancements, with solvers becoming very reliable and competent, providing very robust quotes, doing private market making and striving to improve their view of the available onchain liquidity, as well as taking advantage of the structural benefits of batching. All these are largely due to the solver rewards that solvers receive in COW.

This proposal aims to ensure appropriate funding to continue rewarding solvers and hence supporting a competitive solver landscape, with the goal to establish a state-of-the-art diverse collection of solvers running on CoW Protocol and executing the users’ trading intents in the best possible way, while taking full advantage of the unique feature of CoW Protocol, namely batching of trades.

As already mentioned, the projected estimate of CIP-29 proved to be very accurate, lasting for 6 months. However, one should observe that in the past several weeks, the rewards given in the price estimation competition have exceeded the estimated average of 77k COW per week. Specifically, we observed the following (where we used the main solver rewards Dune dashboard to collect these numbers):

Week of January 9, 2024 - January 16, 2024:

  • 109,017 COW for quotes
  • 195,450 COW for performance rewards
  • 111,195 COW for consistency rewards

Week of January 2, 2024 - January 9, 2024:

  • 103,563 COW for quotes
  • 253,885 COW for performance rewards
  • 52760 COW for consistency rewards

Week of December 26, 2023 - January 2, 2024:

  • 98,802 COW for quotes
  • 180,026 COW for performance rewards
  • 126,619 COW for consistency rewards

Week of December 19, 2023 - December 26, 2023:

  • 99,360 COW for quotes
  • 189,049 COW for performance rewards
  • 117,605 COW for consistency rewards

Week of December 12, 2023 - December 19, 2023:

  • 110,331 COW for quotes
  • 130,968 COW for performance rewards
  • 175,677 COW for consistency rewards

Week of December 5, 2023 - December 12, 2023:

  • 100,368 COW for quotes
  • 147,189 COW for performance rewards
  • 159,456 COW for consistency rewards

The above data indicate increased orderflow to the protocol, which is a very positive development, but on the other hand, it shows that if the protocol continues to increase its market share and the number of orders it executes on a weekly basis, the DAO would end up spending more than the average of 77k COW per week for quote rewards. Concretely, for these past 6 weeks, the average quote rewards per week were 103,573 COW, which is 26,573 more COW compared to the original estimate. On top of that, the COW token price (measured in USD) has consistently been about three times as much (or even more) as it was at the time CIP-27 passed.

For the above reasons, we propose to decrease the per order quote reward from 9 COW down to 6 COW, but also introduce a cap of 0.0006 ETH per quote reward. Concretely, this would mean that a quote would be rewarded with an amount equal to

quote_reward = min{0.0006 ETH, 6 COW},

where the comparison of the 0.0006 ETH to the 6 COW would be based on the market prices at the day of the weekly payout.

As a comment, with today’s market prices, the ETH cap would be enforced and the protocol would be spending an average of 57k COW (based on the data from the past six weeks) for quotes, while if we only look at the reduction from 9 to 6 COW, the protocol would spend an average of 69k COW per week for quotes.

Since we are optimistic that COW will continue to perform strongly, we also propose to adjust the solver competition rewards. As a reminder, the soft budget on solver competition rewards, that splits into performance and consistency rewards, is specified in COW and is currently set to 306,645 COW per week. However, the performance rewards are specified in ETH and have an ETH-denominated cap (but are paid in COW). This means that the value of the consistency rewards in ETH fluctuates with the price of COW. As a consequence, when the price of COW relative to ETH increases, solvers are mainly rewarded for consistency and not for their performance in the competition. Since the recent COW price increase, we have indeed observed that the performance rewards have decreased, implying that the consistency (aka participation) rewards have increased. As we believe that performance rewards provide very strong drive for solvers to keep improving their executions, we propose to increase the upper cap of 0.01 ETH, enforced on a per settlement basis as specified in CIP-20, and make it 0.012 ETH, while keeping the lower cap of -0.01 ETH; this effectively makes the cap asymmetric. The reason for not decreasing the lower cap symmetrically is mainly that solvers still do not fully own the submission process, and, since the main goal here is to increase performance rewards, we believe this is a reasonable compromise (until the protocol reaches a future more decentralized phase where solvers own and control their submissions and thus can be held fully accountable for reverts/cancellations). In combination with this, we propose to make the (soft) budget on the solver competition equal to 250k COW, instead of the 306,645 COW that has been until today. This budget is soft in the sense that the mechanism computes performance rewards by ignoring this budget, and only in the case where performance rewards during a week are lower than that budget, it then distributes the remaining budget as consistency rewards to solvers. Given the data on the past 6 weeks, the increase in performance rewards with this change would be at most 20%, which shows that in most weeks there would still be some budget left for consistency rewards. Specifically, there would be, on average, at least 40k COW remaining to be distributed as consistency rewards. Finally, to ensure that the protocol does not overspend on consistency rewards in cases where the COW token significantly overperforms ETH, we further propose to add an absolute cap of 6 ETH on the total amount of consistency rewards that are distributed on a weekly basis. This means that we would have

consistency rewards = max{0, min{250k - performance rewards (in COW), 6 ETH}}.

This amount of 6 ETH currently corresponds to ~50k COW, and is still significantly larger than what CIP-20 intended as rewards for consistency rewards; as a reminder, the goal of CIP-20 was that around 25% of the budget at the time was given as consistency rewards, which corresponded to around 95k COW when the evaluation of COW was around $0.08 and ETH was around $1,500.

To summarize, we propose the following changes:

  • We aim for a weekly rewards budget of 300k-310k COW, which is a reduction of roughly 20% compared to the budget of 383,307 COW that CIP-20 specified. For this reason, we ask the DAO for 8M COW for the next 6 months, estimated to last from Tuesday, February 6, 2024 until Tuesday, July 30, 2024.

  • The quote reward per order is defined as min{0.0006 ETH, 6 COW}.

  • The upper cap on performance rewards per settlement is increased from 0.01 ETH to 0.012 ETH, while the lower cap remains the same and equal to -0.01 ETH.

  • The weekly soft budget on performance and consistency rewards is reduced from 306,307 COW to 250,000 COW, and moreover, there is an absolute cap of 6 ETH applied to the total amount of consistency rewards distributed.

As with previous related CIPs, it is important to note that this budgeting proposal is independent of additional proposals to potentially change the solver reward structure, which can be configured for appropriate spending according to the budget.


Transfer 8M COW from CoW DAO (0xca771eda0c70aa7d053ab1b25004559b918fe662) to the Solver Rewards safe (0xA03be496e67Ec29bC62F01a428683D7F9c204930)


Solver rewards are planned to continue using the same operational structure, carried out from the solver rewards safe. ETH for compensating gas execution costs is estimated to be fully covered by collecting fees from users. It is proposed to allocate 8M COW tokens for a 6 month period of rewards. This represents a yearly budget of 16M COW tokens, 1.6% of the COW token supply, valued at $4,867,200 according to current market price.


Feels odd to argue the reduction on total rewards based on the market price of COW token. Was the total budget for rewards adjusted upward when the token price was at 0.06USD?

The adjustment between consistency and performance rewards is very reasonable. On that note, I think the consistency reward should be consider a “safety net” for solver teams that are developing their infrastructure and consistently aiming to improve their service to the protocol. I wonder if these rewards could be adjusted to reflect this more accurately. There are some legacy solvers whose relevance will continue falling (e.g. Naive, Gnosis_BalancerSOR, Gnosis_Paraswap, …). Cosistency is good, but rather useless if the baseline is consistent and competitive.

That’s a fair point. I agree that, in principle, the CIP itself could “avoid” referencing the COW price at all, and, for example, only point to the overspending in quote rewards to propose an adjustment of the per-quote reward. But, I think that, moving forward, it would be difficult to be completely oblivious to the COW price, especially when it comes to very simplistic mechanisms such as the CIP-27 quote rewards mechanism. Meaning that one would need to come up with a robust mechanism that automatically adjusts the rewards based on the competition (such as CIP-20) in order to avoid having to “manually” adjust the mechanism depending on the orderflow the protocol receives and the COW price. As an example, in a great future where the COW price is $1, then asking the DAO to consistently provide $9 for every single quote that leads to an order getting executed, even if that quote is a meme token getting matched on Uniswap v2, is arguably not reasonable.

This, to some extent, also applies to the consistency rewards part of CIP-20. Of course, such manual corrections “should” also apply in the unfortunate case where the COW price potentially decreases again; I do acknowledge here that there was no adaptation of the mechanism in the past when the price was very low, but this does not mean that a similar strategy should apply as is in the future as well.

Note that the performance rewards of CIP-20 are not really restricted in any way, but actually will be increased according to this proposal. And since the main strength points of solvers are revealed via the performance rewards, we still feel that with consistently high traffic, solvers will have many opportunities to shine and showcase their strengths, which will lead to them getting generously rewarded via the second-price mechanism that was established with CIP-20, and the increased cap proposed in this CIP.

I would still say, though, that if there is strong objection to the adjustments the CIP proposes, further discussion can take place in order to reach some consensus.

In case you haven’t noticed, the CIP is live and open for voting.

Weighing in here, we’ve voted yes to this proposal given the urgency to re-capitalize the solver budget (with its depletion last week). The proposal highlights several changes to the budget and solver reward mechanism that directionally make sense given recent changes to the market environment.

However I think the point @Filius brought up is spot on in terms of this budgeting and reward mechanism being too static. 6mo is (or can be) a very long time in our industry for the environment to drastically change, as we’ve begun to witness again. Trying to estimate COW and ETH prices or orderflow/volume metrics is unreliable and COW DAO will likely need to remain conservative in how it adjusts its COW emissions up/down for solvers given it (or at least I would) prefers stability in its solver ecosystem.

This is not the right time or place to kick this off, but I would welcome a healthy discussion over the next two quarters on how to evolve the solver reward mechanism to be more dynamic with real-time adjustments based on volume and protocol revenue (vs. static 6mo adjustments based on COW/ETH price). I like @harisang alluding to this as well:

Meaning that one would need to come up with a robust mechanism that automatically adjusts the rewards based on the competition (such as CIP-20) in order to avoid having to “manually” adjust the mechanism depending on the orderflow the protocol receives and the COW price.

Just brainstorming, a couple early thoughts:

  • A base stipend in stables, to offset solvers’ fixed operational costs. There could be a threshold, eg. only during an onboarding period or performance metric (min surplus provided, etc).
  • A performance/volume-driven reward in ETH, given CoW Protocol revenues will likely be collected in this form. May be unnecessary or scaled down if solvers are able to collect their own fees during submission with driver colocation.
  • A performance/volume-driven reward in COW tokens to align long-term incentives between the solver and its continued interest in servicing CoW Protocol
1 Like

Just an update that voting has closed and the CIP has passed.