TL;DR
In a recent post on Ethereum Network Economics, we very briefly referenced MEV, or “Maximum Extractable Value”. This post will offer a quick explanation of that dynamic and take a look at how big of an issue it is today.
What is MEV?
Originally called “Miner Extractable Value” and now referred to as “Maximum Extractable Value”, MEV is essentially just the quantifiable amount of inefficiency still left from transactions on a chain that can be captured by 3rd parties. Primarily, MEV arises as a result of arbitrage opportunities (across DEXs and between DEXs and CEXs), sandwich attacks against DEX traders, and liquidations on lending protocols. Simply, it’s the additional value that can be extracted by selectively ordering/reordering/excluding a set of transactions on the chain.
MEV can simultaneously be viewed as a positive for a network as it improves efficiency, price discovery, and network liquidity, and as a negative, as it can harm users, increase chain congestion, and impose additional costs on the network.
This is not just a theoretical issue, but one with substantial financial implications. Before the Merge (when Ethereum switched from a PoW to a PoS network), estimated MEV extraction nearly reached $700M, a number that is likely a substantial underestimate. Since the move to PoS and the implementation of new middleware such as Flashbots, these amounts have come down somewhat.
Source: MEV extracted via Flashbots
Where does MEV come from?
Historically, most transactions have been submitted to a public mempool for Ethereum, from which miners/validators use to construct the next block. There has been a formalization of the MEV value chain which now includes dedicated searchers (algorithmically create bundles of transactions) and builders (create blocks), which are then passed to block producers (validators) for submission to the network. Block producers can perform any of the functions or outsource any of the earlier steps to 3rd parties. Most MEV extraction can be attributed to the searcher stage of the chain as this is where the actual transactions are selected and ordered.
Source: MEV value chain via Chain.link
At this point, it is likely clear: if a searcher can select which transactions to include in a block, while also including their own, they can pass on a bundle of transactions for inclusion that put their own transactions in position to greatly benefit, such as frontrunning a trade and then backrunning it afterwards. However, there are many competitive searchers operating at the same time, which leads to competitive auctions to have their bundles of transactions included. This value ultimately flows to the block producers in the form of higher priority fees paid to the producers by the searchers. In a long-term equilibrium when on-chain markets are fully efficient, competition suggests that the entirety of MEV value capture will be done so by the block producers (validators) on Ethereum, and thus largely by the token stakers. Why? Because MEV bots will operate up to the point where the revenue they can earn matches how much they have to pay for bundle inclusion (i.e. MR=MC), thus leading extractable profit to 0.
As we showed above, the markets on Ethereum were far less efficient throughout 2020-2022, allowing for hundreds of millions of economic value to be extracted from the network. Today, this number is far lower but still remains. Exact quantification is very difficult, but some groups such as EigenPhi provide estimates. These too are likely underestimates of the size of the actual MEV.
Over the past 30 days, there has been an estimated $2.375M extracted in profit by MEV bots, 98% of which can be attributed to arbitrage strategies. Liquidation MEV is the least profitable strategy, contributing just $3.7K of MEV profit, and though often cited as a major downside of MEV, sandwich attacks accounted for just $49K of that (2.1%). If such a low share, why do sandwich attacks get such a bad wrap?
Source: Profit and transaction volume by MEV type, via EigenPhi
There are two main reasons mainly. They are purely extractive, as they involve the searcher submitting transactions on either side of a user’s intention in order to give them worse execution, and victims can suffer substantial losses at times. Secondly, the volumes attributable to sandwich attacks are massive. To extract that $49K of profit, bots had to conduct $10.63B of transactions. By comparison, the $2.3M of profit captured by arbitrageurs required far less volume to do so, totaling $2.92B of transactions.
Note: This is also a reason why evaluating chain or Dapp traction can be so problematic. The trade volumes, transactions, and user counts can all be heavily influenced by the activity of bots resulting in highly skewed and uninformative metrics.
The vast majority of sandwich attacks result in profits of less than $1.00. In most cases, bots are playing a volume game here with incredibly narrow margins. Outside of 1 outlier, over the past 30 days, the highest-earning bots made just a few thousand in profit. In aggregate, 126,512 transactions were required to generate that $49K in profit.
Source: Economics of recent sandwich attacks via EigenPhi
However, the chance for outsized earnings on a single opportunity can make it lucrative. For example, on October 9, one attack was able to earn $31.5K profit in a single block, through a complicated series of transactions that ultimately netted the bot profits of 23.35 MKR tokens (each valued at ~$1400), at a cost of just 0.07 ETH (~$175). The image below reflects all of the trades that happened during this attack, to get an idea of the level of sophistication of these bots. This all happens in a single block.
Source: Profitable sandwich attack breakdown via EigenPhi
Arbitrage strategies are far more lucrative, potentially generating $50K or more in profit in a single transaction. However, it is difficult to argue that these activities are not extractive of the network (compared to sandwich attacks). The additional liquidity and price discovery provided by arbitrage bots serve to improve the overall experience of transacting digital assets but at the cost of higher congestion and thus transaction costs on the mainnet. Naturally, as interoperability is better increased, and liquidity is less fragmented, remaining arbitrage opportunities will also become more competitive which in turn should shift more of that remaining value capture up to the validators as well while (hopefully) serving to reduce execution costs on the network. This transfer of value to validators serves a second useful function, supplementing a validator’s earnings from network emissions with added fees earned from priority payments. This helps contribute to the long-run security of the network, turning current inefficiencies into long-term strength.
Conclusion
MEV is the embodiment of remaining inefficiencies on a blockchain network. Typically, it is viewed as a net negative dynamic, largely an extractive issue to be eliminated. In the case of sandwich attacks, this is very much the case, but thankfully despite the share of attention they receive, sandwich attacks account for a small fraction of the overall MEV extraction on a network. In the long run, competitive dynamics suggest that more and more MEV value should ultimately accrue to block producers (validators) in the form of priority fees and tips, thus benefiting both Ethereum’s security budget and those holding and staking ETH.
