Ethereum is Dead (for the 60th time)
MEV, Re-orgs, EIP-1559, Layer 2 scaling, wen ETH 2.0?
Unless you’ve been living under a block for the past couple of weeks, you’ve heard all of these terms in some way or another. There have been a number of amazing pieces published in the past month on the topic as well — this will be a bit light on the high level details in lieu of that — so if you’re interested in a more comprehensive backstory I’d recommend you read Dragonfly’s research on the existing MEV landscape here.
Imo, the evolution of MEV poses a large challenge and opportunity to Ethereum, and more broadly applications like DeFi, as we try to maintain decentralization with consensus stability and an effective user experience
This piece will address three main topics which are increasingly becoming more relevant that have not been the focus of most discussions so far:
- MEV in a post-1559 world but pre-merge (August 5th — ~early 2022)
- Layer 2 Rollup MEV (Optimism, Arbitrum)
- Multi-chain thesis driven by MEV
But first, so we’re on the same page… let’s generalize MEV
Let’s think about MEV broadly arising from a window of time where you are not 100% in control of the action you want completed. On Ethereum, that’s easy — when you submit a tx with some gas price and limit it goes to the mempool to be picked up by miners and you are largely no longer in control of where (or if) it gets added into a block. On ETH rollups like Optimism or Arbitrum, you give up discretion once your tx gets submitted to the centralized sequencer/rollup operator.
The window between submitting your transaction and its confirmation is an exploitable opportunity for somebody else, and that value of that opportunity is MEV.
There are three simplistic buckets that MEV comes in: MEV that probably shouldn’t be called MEV because it’s built into the protocol itself (to avoid triggering people over a redefinition, we’ll call it good MEV), unfortunate MEV, and potentially disastrous MEV.
- Examples of ambiguously termed/good MEV: protocols rely on some degree of MEV capture, like liquidations on Aave, Maker, Compound, etc, or keeping market efficient through arbitrage between Uni, Sushi, etc.
- Examples of unfortunate MEV: frontrunning, sandwiching, etc.
- Example of disastrous MEV: consensus threat through re-orgs, time-bandit attacks — always possible in some form if we don’t have probabilistic finality enforced
For the rest of the discussion, I’m going to remove “good MEV” from the discussion and focus on the last two.
The best we can do to limit the large “unfortunate MEV” bucket as an ETH user, apart from resorting to using private relay networks like Flashbots/Archerswap/MistX or going to Request-for-quote RFQ systems (ZRX, Hashflow, Cowswap) is to set a maximum allowable exploitation which DEX’s have termed “slippage.”
Are these good enough?
Not at the moment. Private relay systems running Flashbots’ MEV-Geth are not permissionless (miners have to be whitelisted in order to be forwarded groups of transactions, called bundles). It’s not really private either, since bundles can still be seen by miners prior to inclusion on-chain which can be exploited just the same. While we haven’t really seen this yet, possible in a post-1559 world that this becomes as issue and miners will have to be banned for this. In fact, some searchers will opt to submit transactions through other pools A funny example from this morning: a cryptopunk was mistakenly sold for sub $0.01, where the buyer submitted a transaction through Flashbots with a miner tip of 22 ETH at 99 wei to ensure the Flashbots miner accepted the transaction. Big win for the searcher, miner, unfortunately loss for the typical user.
If you’re interested in learning more about the future of Flashbots, a truly amazing community of white hats, check out Robert Miller’s research proposal around MEV-SGX here for the future of private relays. On the other hand, RFQ systems, effectively implementing OTC trades with DEX’s, assume a fluid market of market makers to deliver quotes and most importantly are only applicable to forms of MEV generated through trading.
Worth mentioning that “unfortunate MEV” is not something new at all to financial markets — if anyone has read Michael Lewis’ book Flash Boys (the inspiration for Flashbots), the same things have been happening in trad markets with HFT: great papers here and here if you’re interested which conclude that “These high-frequency traders are paying for expensive services like colocation and direct data feeds that give them an information and speed advantage that can be used to front-run the trades of institutional investors. By jumping in front of these orders, high frequency traders enact a pseudo-tax on other investors without providing any benefit in return.” Sounds awfully similar to our unfortunate MEV bucket, except in our case it’s a function of code is law, and can’t be regulated by the SEC!
As far as the disastrous re-orgs go, would recommend catching up on the story behind them here, and finally Paradigm + Vitalik on re-orgs after the merge.
Bottom-line is while the community has rallied around multiple reasonable patches to “unfortunate MEV” which are already seeing adoption, we need a better solution and there still remains the question of “disastrous MEV” until the merge
Onto the new stuff… 1) theories for how MEV will look like after London
Most people are zeroing in on EIP-1559 for the upcoming London hard fork tomorrow. Let’s break down 1559 really quickly.
- Instead of a single gas price each tx will now include a “tip” and a “fee cap”
- Tx’s will only be included in a block if its fee cap is at least the block’s base fee
The math then looks like this, where g = gas limit, 𝛿 = tip, c = fee cap, r = base fee, and p = gas price (only relevant pre-1559):
Currently miners running MEV-Geth (85%+ of hash rate) earn, where 𝛿 is the tip per Flashbots bundle (~.3 ETH/block):
Therefore, the difference in revenue a miner makes post-1559 and pre-1559 will be, assuming gas limit g stays the same, will be roughly:
Miners will likely be making less money than they did pre-1559 since the gas price was so high before it’s implementation. While we haven’t factored in what Flashbots bundle tips could be looking like post-1559 (and it’s possible that the delta between those tips increases substantially) it’s highly unlikely to change the outcome since any increased volume through Flashbots would reduce the base fee r as well… which also means that less ETH than expected gets burned.
At the risk of repeating a lot of what was said previously in the original Flashbots piece on the issue here, I think that at the very least we should:
1. Expect L1 MEV capture to become a lot more competitive as miners become squeezed for profits
2. Under EIP-1559, you can’t submit 0 gas fee transactions — this means that there could be more uncled blocks arising from the necessity to pay the basefee in order to be included
3. Temper expectations in general for burns from EIP-1559 which was designed in a pre-MEV world; Flashbots bundles pay miners through smart contracts and as intended, play a role in reducing the base fee on L1 along with the introduction of L2’s, the next topic
Ok enough on L1 ETH… switching gears to the emerging L2 landscape
My favorite dashboard to look at recently has been the MEV guzzler Uniswap v3 deployment on ETH vs. Optimism here. There’s only $7M in TVL on Optimism right now, but I don’t think that anyone’s expecting it to trail much further behind the $2B TVL on the L1 for too long with 1/10 the cost (in the Optimism Alpha there is just ETH, DAI, SNX, USDT, and WBTC).
So why does this matter? Follow the money trail, and we end up with MEV on L2’s becoming THE biggest friction point for DeFi adoption, both from L1 to L2 and within L2 solutions themselves
On L1, MEV extractors are miners — but on rollup L2’s like Optimism or Arbitrum, the ordering power is simply transferred to the rollup sequencer. This sequencer (generalized as a rollup operator) can arbitrarily manipulate ordering just like miners would on L1. At launch, both will have a single sequencer run by the respective teams, with the idea being that the centralized sequencer will be kept in check by a set of validators, very similar to BSC. If validators discover malicious behavior they can contest the invalid state root by submitting a fraud proof which cancels any incorrect transactions and forces the sequencer to forfeit its deposit. For this reason, an arbitrary 7 day dispute window is established to withdraw funds off L2 — however, if there is substantial MEV to be gained on the L2, this could technically be delayed for a long time.
Assuming these centralized sequencers run by Optimism and Offchain Labs are altruistic and don’t pursue MEV themselves, transactions should be processed purely through latency and the time at which the sequencer picked them up. However, this is understandably a big imposition to the permissionless ethos of Ethereum, so both protocols proposed targeting strong L2 fairness despite weaker L1 fairness with Optimism’s MEV Auctions (MEVA) and Arbitrum’s Fair Sequencing Service (FSS).
- MEVA is pretty simple in theory: the right to sequence is a valuable commodity, so the best way to value it is simply by offering the capability to reorder transactions as an auction and give it to the highest bidder. Similar to how mining creates a revenue stream for miners to largely keep the network in consensus, MEVA allows rollup sequencers to monetize their contributions and economically secure the chain in a centralized winner-takes-all manner.
Does this really make sense though? I’m far from the first skeptic of the model (pmcgoohan has a great read here) but the bottom-line is that this still doesn’t fix any of the perverse economic incentives behind unfortunate and disastrous MEV.
MEVA is essentially a completed version of 0xbunny’s Request for Reorg but only applicable to the N-block window on L2’s instead of L1. We legalize frontrunning in the code and in return tap the lawbreakers on the wrist by making them pay some de minimis amount of their earnings back to sequencers who are by design required to be complicit. It’s really setup like a corrupt nation state where the rich get richer at the expense of the users. Plz don’t tell the SEC.
Let’s take a look at Arbitrum’s eventual solution, FSS. FSS uses a Chainlink Decentralized Oracle Network (DON) to first propogate tx’s to a network which then forwards tx’s to the designated smart contract if they are “FSS-enabled”. The oracle network forwards these encrypted tx’s by time of arrival through a design called Aequitas (meaning those with low latency will win) and they cannot be decrypted or observed until they are ordered.
Decentralization is then effectively offloaded on Chainlink oracles, but that’s really the crux of what they are being used for right now in DeFi — to me, this feels like a very natural next step in what oracles are used for in the stack. This makes a lot of sense, and I hope it works.
To summarize:
1. L2 Optimistic rollup solutions bring a level of usability to ETH that haven’t been seen yet, but at inception we place a lot of trust in centralized sequencers run by the teams in extracting MEV (effectively functioning like BSC)
2. We should prepare for possibility of MEV being rampant on L2 rollups once these get turned off — solutions like MEVA and FSS won’t ship at launch and I don’t mean to bash in its infancy before implementation, but it’s important to understand how the game will be changing over the next year.
3. I think bidding wars and consequent collusion around MEVA will become a big theme on Optimism and on Arbitrum it will be a technological race to most effectively submit transactions with lowest latency to the DON (should eventually be much harder to actually extract MEV on Arbitrum)
4. We buidl bigger and better things! Excited for zkRollups solving many of the problems with their Optimistic counterparts.
To wrap up I want to touch on a thesis that many in the community earlier this year were very resistant to — the fact that cross-chain DeFi is more likely than not an inevitability.
MEV proves that blockspace is a premium and value can be extracted from it: multi-chain (either multi-L1, multiple ETH rollups, etc.) is just a question of time at scale with so many different approaches to balance MEV minimization and democratization while retaining some form of decentralization
What are some other L1’s to look at as it pertains to DeFi? SOL and AVAX come to mind as minimizing base potential for MEV. More on these in the future.
Thanks to Blake Richardson, Avi Felman, and Steve Lee for reviewing, BlockTower team on general feedback (follow our new Medium page!), and countless other community members for their research.
You can reach me on my telegram handle @vikruna.
Disclaimers here.
