Burn incentives in MEV pricing auctions
Thanks to Barnabé Monnot, Thomas Thiery and Caspar Schwarz-Schilling for feedback and comments.
Introduction
Overview
This post presents a rudimentary review of incentives for burning MEV under the “simple” MEV burn mechanism presented by Justin, as well as its slot auction counterpart, “execution auctions” presented by Barnabé. The analysis is also applicable to Francesco’s original MEV smoothing design. These auctions—involving builders bidding, attesters enforcing a base fee floor, and proposers selecting a winning bid—will be defined as MEV pricing auctions (in the author’s view, the “execution auction” moniker could also be extended to cover all MEV pricing auctions).
The post highlights how incentives to drive up the price floor (and thus burn more MEV) can emerge in these designs regardless of any direct profit motive among builders for doing so. Importantly, stakers and staking service providers wish to ensure that competitors do not attain more rewards for selling MEV capture rights than them. They may therefore integrate with builders to bid away competing stakers’ profits. Auctions that set a price floor on proposers’ MEV capture rights will thus be influenced by the overarching staking metagame. It is only at this layer that griefing attacks against proposers to burn their MEV capture rights can be understood. Adverse competition during the consensus formation process might hypothetically lead attesters to bias their MEV base fee floor during split views, rejecting or admitting blocks depending on how it impacts their bottom line (in their roles as both builders and stakers). This is something to be attentive to. Naturally, burning MEV might also be considered a public good, and such incentives are reviewed in the text as well.
MEV pricing auctions
In MEV burn–a simple design, Justin formulated an add-on to enshrined proposer–builder separation (ePBS), modifying the MEV smoothing design. Builders can specify a base fee and a tip in their block bids. At some specific time before the slot begins (e.g., 2 seconds), attesters observe the highest base fee among the bids (“observation deadline”) and impose it as a subjective base fee floor when attesting to the proposer’s block. Only bids with a base fee above the floor are accepted, and the base fee is burned.
If builders bid before the observation deadline with the same timing as today, then the mechanism will burn substantial MEV. Concerns have however been raised over the risk of collusion between proposers and builders and lack of proper incentivization. A recent write-up on the benefits of the design and MEV burn in general generated similar worries of a stable equilibrium of late bidding.
The design can be further modified to involve auctioning off the rights to the entire slot, 32 slots in advance (“execution auction”). A benefit of this design is the ability to offer long-lived preconfirmations and—hypothetically—the reduced value-in-flight during the auction. The same concerns raised for the block auction design can be applied to the slot auction design, because the beacon proposer might still benefit from colluding with builders to form late-bidding cartels when selecting the execution proposer.
A modified MEV pricing auction, MEV burn with builder kickbacks, attempts to compensate builders for bidding early. That design is not the focus of this post, but incentives and side effects in uncompensated MEV pricing auctions will affect its relevance.
Five burn incentives in MEV pricing auctions
The outlined concerns of late bidding are valid, but it turns out that it is not possible to analyze MEV burn without incorporating stakers as participating agents. In such an analysis, competition for attaining the most yield will—under equilibrium—drive participants to burn each other’s MEV. Other incentives for burning MEV also exist. The analysis starts from the most idealistic public good example in (A) and gradually builds toward a metagame of active collusion to discourage other stakers in (E) (see Figure 1).
Figure 1. Five types of builders potentially burning MEV in MEV pricing auctions: (A) Public good builder, (B) For-profit public good builder, (C) Extortion racket, (D) Staker-initiated griefing, (E) Staker-initiated griefing cartel. The incentives behind (D) are important to understand (indicated by an arrow).
(A) Public good builder
The first example is a builder that dedicates resources to burning MEV without a direct profit motive. If Ethereum’s users believe that burning MEV is a public good, and in particular if no other incentive is sufficient, they may come together to fund the development and operation of a public good builder. Initiatives to fund public goods are fairly prevalent within the Ethereum ecosystem. The public good builder can for example consistently bid according to guaranteed MEV at the observation deadline in the block auction design. This ensures that the MEV is burned while the builder will not suffer any direct losses from the bid. In the slot auction design, the builder would instead need to bid according to its expected MEV for the entire slot and might bid slightly below to stay safe.
The public good builder will likely not be the best and will often be outbid in terms of tips from other builders in the proposer auction (taking place after the observation deadline), in which the proposer selects a winning bid. But the operation can still be very impactful. After all, priority fees are a significant portion of all value (in this post these fees are also treated as MEV), and some further “low-hanging MEV fruits” are potentially available without dedicating too large resources for extraction. While the builder may use any public goods funding received diligently and not strive for any profit, pursuing an idealistic path can still raise the originators’ public profile and provide significant economic benefits in the future (perhaps not even directly related to building blocks).
(B) For-profit public good builder
A builder that positions itself as providing a public good may also enjoy direct economic benefits from its operation if some validators sympathize with the mission. There may for example be a market fit for builders that do not censor, nor extract various types of toxic MEV. In the block auction design, the builder could keep the MEV base fee in line with the available (non-censorship/non-toxic) MEV during the attester auction, and then pivot to tipping afterward, retaining some small profit margin. The MEV in some blocks is not particularly geared towards specialized searchers, and stakers may not lose that much in tips for some blocks by selecting the public good builder. Therefore, the public good builder could have higher profit margins in the blocks it does eventually get to build than builders that have not positioned themselves as providing a public good. A builder bidding before the observation deadline might of course also hope that its bids are the only ones to reach the proposer in times of degraded network conditions.
(C) Extortion racket
Given the lower effort required for extracting some of the MEV, it seems like (A) and (B) could have a natural position and high impact within the Ethereum ecosystem. But it may very well be that no successful public good builder can be sustained over the long run. After all, many stakers will not be particularly enthusiastic over a builder that burns their MEV opportunities.
Still, consider the importance of a dedicated MEV-burning builder within the staking ecosystem. If the builder is operational, proposers will lose out on a lot of value relative to if it does not operate. Is there a business opportunity here? Perhaps a builder could commit to burning the maximum possible MEV but abstain from doing so if it receives a bribe from the proposer? It seems natural that proposers would be willing to pay for this, since the proposer stands to capture most value from the available MEV if none is burned. But the prospect of competition makes the business model perilous. If a sole extortive builder is profitable, then a few more may try to enter the market as well. There is not much use in paying off two builders if it turns out that a third burned the MEV anyway through a bid. A mechanism for reconciling this ex-post would become rather complex. The validator may then be better off by simply not negotiating with any extortion racket.
While the extortion racket seems unsustainable, it helps to underscore the power that builders have over proposers. The ultimate incentive for burning MEV then emerges when changing the responsible actor from one unaffected by the staking equilibrium (extorting builder) to one that is not (other stakers). The auction will eventually become part of the metagame of the overarching staking equilibrium.
(D) Metagame—staker-initiated griefing
Staking service providers (SSPs) compete for delegated stake and derive income by taking a cut of the staking yield when they pass it back to the delegators. An SSP must ensure that the yield it offers delegating stakers is competitive relative to offers from other SSPs. The MEV pricing auction may therefore lead SSPs to burn competing proposers’ MEV by tightly integrating with builders or running them in-house. If a competitor burns an SSP’s MEV, then the SSP must respond in kind or will lose out on delegators and thus income. When considering the metalevel of SSPs, this equilibrium seems more stable than an equilibrium of late bidding leading to little or no MEV burn. All it takes to break the late-bidding cartel is one defecting SSP builder, forcing others to respond.
An SSP that through a builder griefs other stakers without taking any loss executes something comparable to a discouragement attack with an infinite griefing factor. This is a very advantageous attack, primarily because delegators will flow to the best performing SSP. In addition, a reduction in overall yield for other stakers pushes down the quantity of supplied stake, bringing up the equilibrium yield. Thus, even if some delegators do not flow to the SSP that burns its competitor’s MEV, the expected staking yield (that the SSP will share in the profit from) will still go up, if the competitor’s customers simply stop delegating. Of course, the cost of running the builder must be accounted for. But large SSPs can amortize that cost across a vast amount of yield-bearing validators.
Yet, directly profiting from the MEV is almost always better than burning it. When an SSP’s builder is able to extract more MEV in a competitor’s slot than any other builder, it will still be better off only bidding to a level that ensures it wins the auction. The SSP must thus make a probabilistic judgment as to the uniqueness of its MEV opportunity in the particular slot before deciding how to proceed (or more precisely, any edge in MEV value V_e relative to the second best builder). An SSP builder must in essence bid before the observation deadline up to the point where the expected payoff from burning the marginal MEV is equal to the expected payoff from waiting and hoping to extract it. There are some game-theoretic nuances to this that here will be set aside, with some aspects discussed in the next section. The point is to assert that there are stronger incentives for builders to bid before the observation deadline than what has been previously understood, because a builder might be run by an SSP that indirectly profits from burning other stakers’ potential MEV revenue.
What happens in the metagame to smaller SSPs and solo stakers? They may not afford to run a builder of their own to ensure that their competitors’ MEV is burned. It is of course possible for solo stakers to try to come together to form a union around a builder, where each contributor is guaranteed to see their validators excluded from MEV base fee bids by the specific builder (and receive full tips during the proposer auction). There is then a question of if they will be able to organize such a union, but also if it really would be necessary. On the one hand, if there are several “griefing builders” running concurrently among the largest SSPs, parties holding less stake may not need to run their own griefing builder. Everyone will see their MEV burned anyway, since the big SSPs burn each other’s and everyone else’s MEV. On the other hand, a party not having a griefing builder readily available may be suboptimally positioned when considering the prospect of cartelization.
(E) Metagame—staker-initiated griefing cartel
Can builders operating at the metalevel collude to selectively burn or selectively not burn MEV, depending on the identity of the slot’s validator? The cartel would strive to ensure that all participating SSPs (or any union of solo stakers) receive the MEV in their validators’ proposed blocks, while minimizing MEV in all other validators’ blocks.
However, if attesters are honest, builders can only cartelize to selectively burn or not burn MEV that they uniquely are able to extract. As long as competing builders are operational, this substantially limits the power of any cartel. Therefore, the advantage of (E) over (D) is not substantial.
Proposer is part of the cartel
When the beacon proposer is part of the cartel, members will abstain from bidding before the observation deadline to ensure that as much value as possible flows to the proposer. This type of cartelization has been highlighted as a concern (1, 2) in the debate around MEV pricing auctions. The idea is that participants come to an explicit or implicit agreement to not bid before the observation deadline. Yet the incentive to burn MEV is stronger than previously understood, since stakers outside the cartel will wish to grief cartel members by bidding early (D), and so from this perspective, the risk of late-bidding-cartelization is lower than feared.
It might also be difficult to efficiently uphold cartelization, because it is not possible for members to know which, if any, defected in pursuit of (D). One avenue would be to try to share the profits from every slot to give all participants incentives to hold back bids before the observation deadline. Yet overall, the existence of (A), (B), and (D) means that some value will still reasonably be burned by public good builders or any competitors not part of the cartel.
Proposer is not part of the cartel
When the beacon proposer is outside the cartel, the goal is to deprive it of revenue while still capturing as much of the MEV as possible. It will still be more profitable for the cartel to extract any unique MEV opportunity rather than burn it. Define V_s as the value a builder can attain in the slot auction and V_b as its value for the block auction (from a block built at the observation deadline). When a builder can extract the most MEV, it has an edge V_e over the second-best builder (kept constant for simplicity). Just as in (D), the cartel can bid up to V_b-V_e or V_s-V_e, with the difference that V_e expands if the cartel collectively gains a larger edge against the best builder outside of the cartel. This expansion is what the cartel tries to capitalize on, both when the proposer is part of the cartel (expanding V_e to lower the burn) and when not (expanding V_e to increase builder profits). A challenge—just as in (D)—is that the cartel might not be able to properly estimate V_e. After the observation deadline, the cartel attempts to extract as much value as possible, leaving the MEV either burned or in their hands.
Collusion at other levels
The presentation so far has been somewhat simplistic. It bears mentioning that collusion need not happen at the level of the builders, but can for example happen at the level of searchers or any out-of-protocol relay that the cartel still finds beneficial to maintain before posting to the P2P layer. In all scenarios of successful cartelization, if some stakers (for example solo stakers) are unable to act collectively, they may end up at the short end of the discouragement dynamic.
Risks associated with attester–builder integration
The analysis so far indicates that (D) may have a significant effect on its own but that it does not necessarily lead to the riskier cartelization in (E). But what might happen when we give SSPs tools for depriving each other of revenue? While SSPs will always compete, competition in MEV pricing auctions is on the verge of seeping into the consensus formation process. At the consensus level, all participants are expected to behave honestly and are rewarded for good behaviour. Through staker–builder integration in (D)-(E), SSPs will come to actively influence each other’s rewards, cooperating or griefing each other. A risk is that SSPs might navigate down perilous paths in this landscape.
It has been noted that MEV pricing auctions suffer from attesters potentially having split views of the MEV base fee floor. Biasing the outcome in a split view one way or the other might benefit one builder over another, result in a block being forked out to deprive the beacon proposer of all rewards, or allow the proposer to reap higher rewards when selling MEV capture rights. One concern is that SSPs might eventually try to profit by tuning their attestations of the MEV base fee floor to produce favorable outcomes. This can also be done as part of a cartel. The honest majority assumption need not be broken to derive profits, due to split views. It is only necessary to put a thumb on the scale, and a competitive consensus formation might make such behavior more likely.
Of course, stakers who do not honestly attest to which bids they have observed at which specific time point subject themselves to risks of social slashing if malicious behavior can be uncovered. This is always a potential final resort under proof of stake. In essence, just as it is prudent to be cautious of MEV or excessive issuance as strata for cartelization, it also seems prudent to be cautious of MEV pricing auctions as a stratum for consensus adversity.
Block vs. slot auctions in terms of MEV pricing
Will block auctions or slot auctions burn more MEV? Is one more centralizing than the other? These questions are not easy to answer, because it depends on which burn incentive that comes to dominate, the likelihood of cartelization under different designs, etc. This section will discuss some differences (previous writings on block vs. slot auctions provide a broader perspective).
Block vs. slot auctions concerning (D)
Assume that (D) becomes an important incentive for burning MEV. Further, assume a competitive market without cartelization and perfect information about how much MEV each participant can extract. In the block auction design, the builder can bid V_b-V_e for the block at the observation deadline to maximize burn while retaining opportunities to extract value. It then updates its block and bid through tips in the proposer auction up until the slot boundary. There is V_s-V_b worth of value that the proposer hopes to attain through tips, and V_e worth of value left for the builder (under these simplified conditions).
In the slot auction design, the builder can instead bid V_s-V_e already at the observation deadline. It is just buying the rights to build the block, not committing to its content, and that value is an entire slot’s worth of MEV. Naturally, V_s will here just be an estimate, and the risk that builders take on by bidding on an expected value instead of a tangible value might be worth some fraction of the total bid value. But incomplete information around competitors’ eventual final bids will likely serve to pull down the bid value at the observation deadline more. The staker–builder can ideally burn V_s-V_e of a competing beacon proposer’s auctionable MEV, and again retain V_e for itself. The difference in MEV burn between the two designs is then V_s-V_b.
If the staker–builder could estimate V_s also in the block auction design (which nominally is easier since it bids much closer to the deadline), it could bid V_s-V_e-V_g already at the observation deadline. Since the bid is attached to a block containing only V_b of MEV, V_g is reserved as a tip for the proposer auction. If there is no tip, the proposer might elect to pick the block from the observation deadline, depriving the builder of V_s-V_b. However, while the proposer might specifically wish to do so if the same builder bids with low tips also in the proposer auction, a staker can obfuscate its identity by running several builders (the kickback design disincentivizes obfuscation).
In either design, it seems most likely that the burn ends up being lower than these theoretical maxima due to incomplete information in combination with the fact that capturing the MEV is more valuable than burning it. The staker–builder will therefore operate with quite some margin to maximize expected profits.
Block vs. slot auctions concerning (A)-(B)
The analysis for (D) is to some extent also applicable for (A) and (B). The public good builder could theoretically bid higher in the slot auction than in the block auction. However, the risk associated with overbidding in the slot auction design might be more serious for these builders. In the block auction design, the available value will be much clearer, making it easier for an unsophisticated builder to make low-risk bids.
Value of preconfirmations
As previously mentioned, the slot auction design facilitates execution layer preconfirmations, which can provide a welfare gain to Ethereum. In addition, their value can be burnt (just as in execution tickets), since builders are bidding to attain that value. This increases the burn of the slot auction design.
Builder centralization under competition over expected MEV
If builders have different strengths and weaknesses, they will intermittently attain the highest V_b in the block auction design. While one builder might be able to extract the highest MEV in expectation, not all blocks will play to its strengths. However, in the slot auction, builders bid on expected MEV, and one specific builder might then always have the highest expected V_s. This could potentially be a centralizing force, depending on how secondary markets evolve.
Conclusion
There are strong incentives for burning MEV even in designs that do not directly compensate for it, for example to provide a public good service or to ensure that other participants in the staking metagame do not attain a higher yield. Uncompensated MEV pricing auctions accommodates these incentives. Of particular relevance is staker-initiated griefing (D). It seems clear that SSPs will seek to influence builders’ bidding strategies, and this can lead to staker–builder integration. Still, this form of integration does not necessarily lead to censorship or higher MEV profits; thus not negating sought benefits of proposer–builder separation. If it is desirable to give an outside party an independent incentive to burn MEV, then builder kickbacks are an option. They can also be applied to the slot auction design.
When implementing a MEV burn mechanism, it is important to ensure that the burn mechanism does not accidentally set fire to Ethereum’s consensus mechanism. Giving SSPs tools for griefing each other could lead to adverse competition during the consensus formation process. A particular concern is then if emerging attester–builder integration leads attesters to bias their MEV base fee floor, rejecting or admitting blocks depending on how it impacts their bottom line (in their roles as both builders and stakers). Which of the different scenarios (A-E) that would predominate is seemingly a more important parameter when evaluating the merits of MEV pricing auctions than the mechanism’s ability to burn substantial MEV (which this post suggests it can).