Attester Questions and Incentives
As mentioned by @Mister-Meeseeks, I think we need more specific details regarding the attesters. Some of this might already be answered and I’m just missing the details from somewhere.
As we obviously know, block production is an infinite game. Attesters are validators, and thus rationally would like to maximize their rewards in both attesting and proposing (while minimizing slashing). Where possible, rational attesters would want to establish a minimal payload base fee precedent so this behavior is reciprocated during their turn in proposing. If possible, clawing back rewards from being burnt is a better financial outcome for validators. Because of this incentive (and some others to follow later in this response), I am curious about the following questions:
- Is there punishment for any attester misbehavior? Under what circumstances is it considered misbehavior, and what is the punishment?
- Are attesters required to broadcast their local base fee floor by some deadline?
- Can attesters broadcast their local base fee floor and then update it prior to some deadline?
- Is there some point in which attesters are absolutely committed to a base fee floor prior to their attestation, or is this just a soft definition that they can individually update prior to the proposal attestation?
Without these answers, it’s a little harder to wargame the nuanced incentives each party has. I’ll continue on with some of my other points, but many of them may be subject to how the above questions are answered.
Potentially Overgenerous Assumptions
Information
This defection strategy is assuming the builder has more information than they would in reality:
- Builders do not know with certainty the total extractable value each other bidder has while the auction is ongoing.
- Due to cross-domain MEV (ex: CEX-DEX arbs), builders do not know with certainty the total extractable value they themselves will have, but the certainty increases as time passes.
With the potential uncertainties above, a reasonably confident competitive builder should attempt to not defect, as they still have the opportunity to be the winner. What’s left are very unconfident builders, who almost by definition are likely to not have as much value to be captured.
Latency
This defection strategy to win the auction also assumes some timing games are a certainty. For simplicity, assume latency includes processing time for the endpoint.
(D + excess\_proposal\_window) < (latency_{defection\_to\_competitor} + latency_{competitor\_to\_proposer})
sidenote: adjust the left side of the inequality depending upon how Attester Questions are answered.
If this latency assumption is violated, the non-defectors can still update their bids to include the expected base fee floor (as mentioned by @jasalper). Thus, the probability of winning selection is only minimally increased and the incentive to defect is minimized.
Synthesis
If both the information and latency assumptions are violated, there exists a very large incentives gap between honest and rational builders. Honest builders will earn little to nothing (many of which may discontinue operations), and rational builders will continue to persist by their profitability.
Structured Bids
There is a possibility that the latency assumption doesn’t need to be violated in order for a rational builder to minimize burn for the benefit of themselves and the proposer alike. Consider this example, under the assumption that bids are not required to be observed by attesters prior to proposal:
An independent relay is constructed, similar to what is shown in the graphic below from Relays in a post-ePBS world. As seen in the graphic, the bids from relays to proposer are not by default visible to outside parties such as attesters.
Because of mev-burn, builder B likely wants to keep their bids hidden from attesters unless the relay is unreliable, or they want to defect. In order to mitigate being usurped by a last-minute defection from other bidders, the relay announces that it will accept and relay structured bids. When the relay submits bids to the proposer, it just submits multiple bids to counter other defections only when necessary. Here’s roughly what a simplified structured bid could look like:
Each builder’s actual block contents are functionally the same, with the exception of varied payments. Notice that the proposer is still incentivized to select the lowest payload base fee bid possible, and both proposer and builder stand to benefit. Of course, the marginal changes in rewards don’t have to be split 50/50. As seen in the following table, both proposer-favored and builder-favored environments can still result in bid structures that incentivize minimal payload base fee for the benefit of both parties.
There are some additional configuration details for relays that would still need to be finalized, but this model appears possible to minimize payload base fee more than intended by design of mev-burn. Note that the relay only needs to anticipate what the attesters will require for the base fee floor – not necessarily what the highest payload base fee in a bid is. Note that this relay design is increasingly more useful the more centralized the builder set is. Today there are very few highly skilled builders, thus we can likely assume that they will be attracted to solutions such as this (or similar via vertically integrated builder/relay).
Historic Data Does Not Necessarily Apply to a New System
As also mentioned by @tripoli, the assumption that past data applies to this new proposed system seems highly flawed. You’re completely altering an incentive mechanism, therefore you should expect observant and rational agents to behave according to the new system – not the old one.
Assuming honest attesters, mev-burn signals that it will attempt to burn the majority of value observed from all bids D seconds prior to the beginning of the slot. Unless you as a builder are nearly certain that you will lose the auction before it has concluded (see Potentially Overgenerous Assumptions), there is no reason to defect before D. And even if you choose to defect before D, your defection bid’s success is still contingent on the Latency assumption. Obviously your defection bid does force additional burn, but we can likely assume that its direct value add to you is de minimis. The most convincing argument to defect (when you believe you will lose the auction) is forcing burn deprives your competition of future capital that they could use to improve themselves.
That’s a lot of conditions to rely on, and depending upon how the Attester Questions and Incentives section is addressed, may further layer on more conditions surrounding what the payload base fee floor looks like in reality.