Domothy and I co-authored this article. PBS is still an active area of research, but this comprehensive post aims to aggregate and summarize where research is so far- and where it’s headed.
In the existing architecture, block proposers (previously miners) select transactions in the mempool to build a block. On Ethereum, these consist of validators running a consensus and execution client in tandem. Theoretically, a block builder would select transactions paying the highest fee- but this is not a firm requirement. Because a builder has complete control over transaction ordering and inclusion, builders can run complex MEV strategies. MEV is highly prevalent wherever there is activity, and as a result, Ethereum is the most lucrative battleground, with over $675M extracted to date.
Technically speaking, an Ethereum transaction is simply a series of bytes. Therefore, the costs involved in creating and sending it are negligible. A good example is on-chain arbitrage. When the same asset is priced differently between exchanges, it creates an opportunity for value extraction. Any money made from this transaction, minus gas fees, is profit essentially extracted from the economic activity on the blockchain.
However, there is a catch: Due to the transparency of the blockchain, anyone with a node can identify and submit the same transaction. No matter how many people send their transaction, there will only be one winner: the one who has their transaction confirmed. As you might imagine, this greatly benefits the person who decides which transactions go in the next block and in what order. They can simply slip their transaction through!
This example illustrates what is known as an “MEV opportunity.” Given the list of pending transactions waiting to be confirmed, and the current global state of Ethereum, there exists a specific order of transactions that will let the block builder extract the most value out of the ecosystem. Most of the time, MEV opportunities are more complex and require multiple transactions in just the right order, but the principle is the same: The block builder has a massive advantage.
Several years ago, when Ethereum was still secured by Proof of Work, the concept of MEV was mostly theoretical. Instead, transactions were simple: “Here’s what I want to do, and here’s how much I’m willing to give to the miner for including it in the next block.”
The miner would simply go down the list of pending transactions, order them by highest fee, and fit as many as possible inside a block. The intricacies of the transactions themselves were irrelevant. This satisfied some notion of fairness: Every miner would run the same code, play by the same rules, and make the same amount of money (relative to hashrate).
But simply put, this couldn’t last. Financial incentivization breeds competition. The MEV transaction that offers the most for inclusion wins, as the miner is also financially incentivized to pick the most profitable transactions. In these scenarios, the miner is the one who profited the most without even playing. Given enough competition, the profits always consolidate with the miners (or validators, in the case of PoS Ethereum).
The first consequence showed itself quickly. Every MEV transaction, regardless of success, still had its attempt included on-chain. This wasted a lot of Ethereum’s already limited blockspace, which resulted in incredibly high fees.
When miners realized the power granted to them as block builders, a new centralizing force emerged. A block builder can guarantee the inclusion of their MEV transaction, regardless of the bid.
Note: after EIP-1559, block builders still have to pay the base fee.
The new strategy was for block builders to actively monitor the chain to look for MEV opportunities to exploit themselves directly. It’s easy to see why this is a strong centralization vector, independent of consensus mechanism: Creating blocks more often will result in more MEV extraction. But, unfortunately, MEV breaks the guiding notion of fairness by consolidating extraction over time due to economies of scale.
This centralization results in a second-order consequence: Beyond ordering transactions in a block, block builders also have the choice to add their block as expected or roll back the chain a few blocks back to effectively “steal” all the MEV opportunities from recent transactions. In some extreme cases, stealing the previous block’s MEV could be the more lucrative choice, which could easily lead to an unstable blockchain that keeps splitting into smaller forks, resulting in longer settlement times.
Thankfully, this last concern never materialized on the base layer. However, relying on a few prominent actors acting in good faith against strong financial incentives is not the level of resilience Ethereum strives for. With a long enough time horizon, MEV is an existential threat to Ethereum if left wholly unchecked.
Regarding MEV, there have been many proposals to mitigate the aforementioned consequences, and it is still an open area of research. Some potential solutions:
Force blocks to be ordered by fee
Back to gas wars
Validators can ignore inconvenient transactions due to a lack of mempool consensus.
Force random order of included transactions
Back to gas wars
Added latency from transaction consensus
Validators could keep blocks artificially empty except for their MEV transactions, which results in a terrible experience.
Reduces toxic MEV
Requires increased validator honesty assumption
Increases censorship resistance
MEV from validator self-inclusion is unaffected
Many more potential solutions, both cryptographic and cryptoeconomic, are being explored. “Solving MEV” as a whole is impossible. Any hurdles would result in more esoteric MEV extraction, exacerbating centralization. Surprisingly, the next best thing is to accept that MEV is inevitable and will always be a centralizing vector whether we like it or not.
The solution is to work around it by separating the process of proposing a block from building it, hence the name Proposer-Builder Separation. The idea of a “block proposer” already exists in Ethereum’s Proof of Stake design. Every 12 seconds, a single validator is elected as a block proposer. It doesn’t matter how the block is built. The validator’s expected job is to propose it to the rest of the network, and the job of every other validator is to verify it to make sure it doesn’t break any rules.
The general idea of PBS is that builders will compete and submit bids to whichever validator is in charge of proposing a block. The only thing this validator has to do is propose whichever block offers the highest bid.
With PBS, the highly adversarial MEV battleground is carefully isolated to where the block building takes place. In addition, the auction system allows all validators to easily benefit from MEV opportunities found by block builders, regardless of scale.
It might seem unintuitive to give up and turn the whole thing into one big auction without actually trying to fix MEV itself; however, MEV is simply too lucrative to dismiss as some rare edge case. If we don’t enshrine such an auction directly in the protocol, it will still exist in some form, except it will be out-of-band, opaque, and subject to the centralizing forces mentioned previously. If left unchecked, it’s easy to see what private block-building auctions might look like and all the harm they can potentially cause to the decentralization and credible neutrality of the blockchain.
With PBS, we benefit from this unavoidable economic reality instead- We get to design the rules around what such a market should look like and eliminate the parts we don’t like. We can make it fair, transparent, and readily accessible to every validator.
What should an ideal market look like?
Proposers should have no incentive to prefer certain builders
Builders should have no incentive to prefer certain proposers
Minimal proposer overhead
Simplicity of consensus
Implementing PBS directly in the protocol is complex and will take some time. In the meantime, some out-of-protocol solutions have emerged to achieve the same goals, with some slight trade-offs. While not being enshrined in the protocol officially, out-of-protocol implementations allow for a higher degree of experimentation and serve as an excellent testing ground for final implementation. Flashbots has created a block-building market with all-or-nothing bundles and conditional payments.
Their tools include:
Bad for client diversity
Only worked pre-Merge
Client-agnostic, solving the problem with the previous implementation
Validators cannot manipulate blocks (they receive the block body only after signing the header)
Out-of-protocol markets still have some downsides:
Relays still require trust
A proper PBS design is still an active area of research. We have the guiding principles, but final implementation is not a trivial task.
Builder centralization is less harmful than proposer centralization, but still not ideal. As a result, there is ongoing research for eventually creating a network of decentralized builders with the hopes that it may outperform centralized entities.
Implementing crLists still allows for censorship, but long-term transaction censorship becomes prohibitively expensive in a competitive market, requiring builders to fill blocks. If the builder does not fill the block, the space is filled with proposer-selected transactions. This increased defense vastly reduces the existential threat of extreme builder centralization. Encrypting the mempool can also fix censorship issues and toxic MEV simultaneously but requires a low-level protocol upgrade.
Even if earnings are democratized through PBS, proposer luck is still an incentive for pooling stake. The more validators an entity has, the more likely they are to propose blocks and claim MEV rewards. Implementing MEV smoothing into the core protocol allows us to reach the ideal fair scenario where staking rewards are proportional to stake, as MEV rewards would be aggregated and paid equally to validators. In addition, in-protocol MEV smoothing, if implemented optimally, reduces the incentive to circumvent an in-protocol PBS design.
There are a lot of significant research initiatives in the PBS space, but there is a distinct lack of formal analysis regarding incentives or economics. Moreover, the research space is changing rapidly, and many more developments are inevitable before we reach the analysis and implementation stage.
A lot has happened in MEV and protocol research over the past few years, and the subject broadens the more we learn about it. While we still have open questions, we are on the path to a decentralized, highly resilient, neutral global network. If you want to dive even deeper, Domothy has put together a list for further reading here.
Thanks for reading! We hope this was helpful. Follow both @pseudotheos and @domothy on Twitter to get notified of future posts. This article is licensed under CC BY-SA. Thanks to Jon Charbonneau, Dmitriy Berenzon, and No Clue Capital for their feedback and comments.