Thanks to Barnabé Monnot, Thomas Thiery, and Data Always for their feedback and discussions.

How Proposers Contribute to Credible Neutrality

Proposers have monopoly control over what happens to the block in their slot. Usually, proposers outsource their block construction duties to the highest-paying builder through an auction, relinquishing all input a proposer may have into block construction. Sometimes, however, proposers may want to make a more nuanced decision about what happens with their block.

A proposer may locally build a block for any reason, but many do so to contribute to the network's credible neutrality. Instead of consistently building locally or outsourcing to MEV-Boost, Flashbots introduced a tool allowing proposers to make a more nuanced decision called the min-bid parameter. It enables the proposer to set a threshold value, which is the minimum bid the highest-paying MEV-Boost builder must pay. If the highest-paying bid is above the threshold value, the proposer outsources block construction; if the highest-paying bid is below the threshold value, the proposer locally builds.

The minimum bid parameter allows the proposer to make a more granular choice between two objectives: credible neutrality and profit maximization. As per Data Always’s last analysis, the parameter leads to 3% of blocks reverting to local building, representing about 40% of locally built blocks.

Another tool that enables the proposer to make a more nuanced choice between local building and outsourcing is localBlockValueBoost (it has different names for different clients). This parameter allows the proposer to outsource block construction to a builder if the highest-paying bid is at least a certain percentage higher than the value of a local block. See Figure 1 (taken from Data Always) for an example of LocalBlockValueBoost and min-bid.

LocalBlockValueBoost could be valuable to a proposer if it only wants to outsource block construction if the profit from outsourcing is significantly higher than local block production. This relative comparison of block values is important in the presence of private order flow. In one block, the value of a local block and the value of a builder block could both be 1 ETH, whereas in the next block, the value of a local block could be 0, and the value of a builder block could be 0.08 ETH. With a LocalBlockValueBoost of 10%, the first case would lead to local building and the second to outsourcing, whereas a min-bid of 0.1 ETH would lead to outsourcing in the first case and local building in the second.

LocalBlockValueBoost seems potent in theory, but it is still limited in practice. Data Always found that setting it at 10% for 100% of proposers would lead to less than 1% of MEV-Boost blocks reverting to local building, thus making a significantly smaller contribution to credible neutrality than the min-bid parameter currently does.

Both min-bid and LocalBlockValueBoost allow the proposer to express some of their preferences over block construction. Similar tools can be further explored and are currently being built. Examples include the out-of-protocol inclusion lists via Commit-Boost and the P2P market proposed in EIP-7732 that could also be used for proposers to express preferences. A proposer may prefer a bid from the P2P over a bid from an RPC builder because the P2P may create more diversity amongst bidders (h/t potuz).

What Changes in ePBS?

How will these preference expression tools fare under ePBS? EIP-7732 introduces a significant change in the market structure compared to MEV-Boost. In its current form, the EIP entails that a proposer must commit to an execution payload six seconds before it must be revealed at the latest; this is a block auction. In the ePBS Breakout Calls, it has been discussed that moving to slot auctions could be beneficial for various reasons. One concern with slot auctions is that proposers would not be able to express their preferences sufficiently well through methods like LocalBlockValueBoost, leading to a decrease in weak censorship resistance: the time to inclusion for a valid transaction that pays sufficient fees but is excluded by certain block producers for exogenous reasons.

LocalBlockValueBoost would not work as well with slot auctions because sophisticated builders bid based on the maximum value that the execution payload could have at the time it must be delivered. In contrast, the proposer must decide between committing to one of these bids and committing to local building six seconds earlier. The highest-paying bid and the value of a locally built block at the time of commitment reflect drastically different information sets: the former is the expectation of the block value to the highest-paying bidder six seconds into the future, whereas the latter is the value of an unsophisticated block currently. The problem is that an unsophisticated proposer cannot accurately predict the local block value six seconds into the future.

In the worst case, this would mean that LocalBlockValueBoost would never lead to a proposer choosing local block construction over outsourcing. Even this worst case would mean that there would be a small decrease in the number of locally built blocks since even with 100% adoption of the current LocalBlockValueBoost, it would only revert 1% of MEV-Boost blocks to local block construction.

Tools that do not rely on local block value would work as well under slot auctions as they do today. The usage of min-bid, for example, would not change if ePBS were to use block or slot auctions. Therefore, the ecosystem retains the preference expression tool that currently contributes the most to credible neutrality.

The Big Picture

The change from block to slot auction ePBS may mean that LocalBlockValueBoost is no longer a viable way for proposers to express their preferences. This may increase the time to inclusion for certain transactions the top builders do not want to include. However, the change from block to slot auctions also has more effects. All of these effects should be jointly considered to obtain a better view of the equilibrium effect on the time to inclusion for different types of transactions.

This section considers the consequences of time to inclusion for all transactions. This is relevant since people may worry that the time to inclusion for certain transactions that top builders do not want to include increases as LocalBlockValueBoost loses efficacy. The following arguments, however, suggest that the average time to inclusion for all transactions decreases with slot auctions. Therefore, the average time to inclusion may also decrease for transactions that some builders do not want to include. See this post for a comprehensive list of arguments in favor and against slot auctions.

First, in block auction ePBS, sophisticated builders may choose not to reveal their execution payload if the value of the execution payload has become negative when the builder must reveal it. Essentially, a sophisticated builder buys an option to reveal their block. This will primarily lead to more missed execution payloads in block auction ePBS than in slot auction ePBS, meaning that the average time to inclusion for all transactions will increase. See this note for more information on this problem.

Moreover, since sophisticated builders are aware of this option not to reveal, their valuations will be strictly higher than those of an unsophisticated proposer who does not use this option. This means that even in block auction ePBS, LocalBlockValueBoost may not work as well as in MEV-Boost currently. This issue stems from the six seconds between a proposer committing to an execution payload and the execution payload being revealed. Qualitatively, this is the same issue that causes LocalBlockValueBoost not to be effective in slot auctions: sophisticated bidders base their bids on a richer information set than unsophisticated proposers. Quantitatively, LocalBlockValueBoost would work better in block auction ePBS as the option to not reveal an execution payload in block auctions will not be used as often as the option to change execution payload contents in slot auctions.

Finally, slot auctions allow a block producer to include transactions during the time interval after a proposer commits to the builder and the builder reveals the execution payload, whereas block auctions do not. This means the average time to inclusion for all transactions is lower in slot auctions than block auctions.

Conclusion

Proposers may want to condition their use of builders on the value that builders bring them. If the proposer builds a block locally, this could help credible neutrality. Slot auctions may make it more difficult for proposers to condition their use of builders on the value that builders provide compared to the value a proposer can generate. This may lead to more proposers outsourcing block construction and hurt credible neutrality. However, slot auctions do not hurt proposers' most potent tool: min-bid. Therefore, losing a tool like LocalBlockValueBoost does not hurt credible neutrality much.

Finally, slot auctions may lead to a slightly increased time to inclusion for transactions that would have been included if proposers had been able to use LocalBlockValueBoost effectively. However, shifting to slot auction ePBS decreases the average time to inclusion for all transactions, and slot auctions will cause fewer missed execution payloads than block auctions. This is beneficial for almost all transactions and helps credible neutrality.