rounded

Original title: It's time to talk about L2 MEV

Original author: sui14

Original translation: Ladyfinger, BlockBeats

 

Introduction

 

In this post, we aim to provide a data-driven overview of the current state of L2. We monitor the importance of the Dencun upgrade’s gas fee reduction for L2 in March, examine how activity on these networks has evolved, and highlight emerging challenges driven by MEV activity. Additionally, we discuss potential barriers to developing MEV tools and solutions for L2.

 

The good: L2 adoption after Dencun upgrade

 

Gas costs dropped 10x

 

Ethereum L2 gas fees consist of two parts: the cost of executing transactions on L2, and the cost of submitting batches of transactions to Ethereum L1. Different L2 gas fee structures and ordering rules vary depending on their development stage and design choices. For example, Arbitrum operates on a first-come, first-served (FCFS) basis, with transactions processed in the order they are received. In contrast, Optimism (OP Mainnet) and Base as part of the OP Stack use a priority gas auction (PGA) model that combines L2 base fees and priority fees. Users can choose to pay higher priority fees in order to be included faster and appear earlier in the block. Understanding the fee structure is critical to understanding the growth of the ecosystem and the MEV dynamics.

 

Historically, Ethereum L1 fees made up the majority of the total fees users had to pay when transacting on L2, accounting for over 80% of the cost, as shown by the black bars in the figure below. However, after the March 14 Dencun upgrade, L2 switched from using calldata to a more economical method, so-called "blobs 1", for submitting batches to L1. This temporary storage contains its own gas auction, consisting of a blob base fee and a priority fee.

 

Data Sources

 

There has been a significant reduction in fees paid by L2 to L1 since Dencun - the chart shows that the gas cost breakdown of the OP Stack chain has changed significantly, with L1 costs plummeting from 90% to just 1%, while L2 costs now account for 1% of the total 99% of cost. This shift resulted in an overall roughly tenfold drop in average total gas fees on L2, with OP Mainnet for example seeing average gas fees plummet from around $0.5 to $0.05 per transaction.

 

Data Sources

 

Activity surges on L2

 

After the cost reduction, there has been a clear increase in activity and usage on L2, which can be seen from the surge in gas fees on L2 in the figure above. Notably, on March 26, Base's average gas fee exceeded its highest level before the upgrade. In order to accommodate more transactions and reduce network congestion, Base increased its gas target starting on March 26 and has made several adjustments since then.

 

The chart below highlights the number of daily transactions on L2, showing the significant growth of networks such as Arbitrum, Base, and OP Mainnet. In particular, Base's daily transaction volume has grown fourfold and now handles approximately 2 million transactions per day.

 

Data Sources

 

While it’s difficult to determine whether this is a result of organic participation or influenced by incentive programs and Sybil activity — since the end of last year, with improving market conditions and the arrival of memecoin season triggered by WIF on Solana, active addresses and DEX trading volume on all major L2s have clearly increased after the EIP-4844 upgrade, especially on Base and Arbitrum.

 

 

Assets flowing to L2

 

With the improvement of market conditions and the arrival of the memecoin season triggered by WIF on Solana, TVL on L2 has continued to rise since the end of last year. Notably, Base has become the fastest growing chain, and recently surpassed OP Mainnet in total TVL.

 

Data Sources

 

Since the beginning of March, Base has seen inflows of around $1.5 billion in USDC, part of which is Coinbase moving customer and corporate funds onto Base. According to Artemis data for 11 major bridges since January 2024, there has been $14 billion in outflows from Ethereum to the main L2s. Arbitrum leads with around $7 billion, followed by zkSync, Base, and OP Mainnet. Further data from Debridge Finance, a cross-chain bridge widely used in EVM chains and Solana, confirms that Arbitrum and Base are the top recipients of all outflows.

 

Data Sources

 

The bad: Hidden MEV activity increases as gas fees decrease

 

As we further inspected the transactions, we noticed that bot transaction activity was driving up gas fees and rollback rates on L2. We will explore this issue more fully in the next section by using Base’s statistics for a case study, highlighting the impact of cheaper gas on L2 after the Dencun upgrade.

 

Dencun’s upgraded L2: Similar to Ethereum without Flashbots, but without the transaction pool

 

Network congestion

 

The challenges began to emerge on March 26, when the average daily gas fee of the Base network briefly surged, surpassing the level before the Dencun upgrade. However, on June 3, Base raised its gas target to 7.5M gas/second, compared to 2.5M gas/second during the Dencun upgrade, which brought the average gas cost back down to about 5 cents.

 

On the Base network, the contracts that consume the most gas include Telegram exchanges BotSigma and Banana Gun, as well as digital wallets and DEXs such as Bitget and Uniswap. In addition, there are many unmarked contracts involved in activities such as token minting, meme coin trading, and atomic arbitrage. These contracts are the top contracts on the Base network ranked by gas fee payment.

 

 

By comparing the behavior of popular Telegram Bots, such as BananaGun, it is clear that the transactions they conduct incur much higher gas fees than normal transactions. After the Dencun upgrade, users using the BananaGun Telegram bot saw gas prices spike to a peak of 30 Gwei when performing transactions on the Base network. Although this rate has since stabilized at around 3 Gwei, it is still 43 times the gas fee that other transactions have to pay.

 

Daily gas prices on Base, comparing Banana Gun transactions to other transactions

 

When analyzing the average monthly gas prices paid by all major DEX trading bots on the Base network and comparing them to non-Telegram bot transactions (represented by the black bar), it is clear that users using trading bots incur significantly higher gas costs. Below is a comparison of monthly gas prices on the Base network, showing the difference between all Telegram bots and other transactions.

 

Data Sources

 

High rollback rate surges

 

The rollback rate of transactions in a blockchain network is an important indicator of its health. We noticed an increase in rollback rates after the Dencun upgrade, especially on L2 networks such as Base, Arbitrum, and OP Mainnet. Currently, the rollback rate of Ethereum Mainnet is about 2%, while the rollback rates of Binance Smart Chain and Polygon are between 5-6%. Before the Dencun upgrade, the rollback rate of Base also remained at about 2%, but then it rose sharply to about 15%, and peaked at 30% on April 4. At the same time, Arbitrum and OP Mainnet have also seen periodic surges in transaction failure rates, which fluctuate between 10% and 20%.

 

Cross-chain transaction rollback rate

 

After further analysis, we found that high rollback rates on the L2 network do not always represent the actual experience of ordinary users. Instead, these rollbacks are likely caused by MEV bots. By adopting the following heuristics (Query 2), we identified a group of router contracts that exhibit bot-like behavior - they show high rollback rates when executing MEV withdrawal transactions:

 

Since Dencun was upgraded,

 

Active Router: This contract has processed more than 1000 transactions.

 

Limited Interaction EOA: Fewer than 10 EOA (Externally Owned Account) wallets have interacted as a transaction sender.

 

Sender distribution: Less than 50% of transaction senders only sent one transaction, indicating that the user population does not exhibit a long-tail distribution. This suggests that the router is unlikely to be used by retail users.

 

Behavioral patterns: Transaction history covering exactly 24 hours or showing multiple transactions within a single block, indicating non-human behavior.

 

· Exchange concentration: More than 75% of successful transactions involve an exchange.

 

Detected MEV transactions: More than 10% of successful transactions used the atomic MEV strategy, as detected by hildobby’s heuristics.

 

Using these criteria, we detected 51 routers on Base, which likely represents a conservative lower bound estimate of Bot activity on Base.

 

We divided all transactions processed by routers on the Base network into two groups and conducted a comparative analysis. The results show that bot-like routers have significantly different rollback rates compared to other transactions: bot-like contracts achieve an average rollback rate of 60%, which is six times higher than the approximately 10% observed for other transactions.

 

Daily rollback rate on Base, by Bot-like contract vs. other transactions

 

Based on the above data, we can infer that automated trading activities such as MEV robots and Telegram robots are likely one of the main reasons for high gas fees and high rollback rates on the Base network.

 

The single sequencer architecture of L2, combined with the lack of a public transaction pool, has fostered a large number of MEV strategies that exploit the sequencer, which have become the main cause of network congestion. This congestion is particularly evident on L2 networks that adopt the priority gas auction (PGA) mechanism, such as OP Mainnet and Base. The result is not only congestion in the network, but also a large amount of block space and gas fees wasted due to rolled back transactions and MEV seeker activity. This is similar to the situation on Ethereum before the emergence of Flashbots, except that due to the lack of transaction pools on L2 at present, there is no sandwich MEV phenomenon.

 

How big is the MEV scale on L2?

 

Understanding MEV activity on L2 networks is critical to assessing its impact. However, there is no widely agreed-upon number for L2 MEV data that is verified by multiple sources and reliable methods. In addition, compared to the Ethereum mainnet, L2 lacks real-time monitoring data provided by tools like mev-inspect, libmev, and eigenphi, which are critical to measuring the total amount of MEV and miners' profits.

 

Some of the L2 MEV datasets and studies published to date include:

 

Open source datasets built by hildobby at Dune Analytics (inspiration links: Sandwich | Sandwich | Atomic Arbitrage)

 

A research paper by Arthur Bagourd and Luca Georges Francois titled “Quantifying MEV On Layer 2 Networks”, which quantifies MEV on Polygon, OP Mainnet, and Arbitrum using the mev-inspect implementation. This research was funded by Flashbots.

 

A research paper, “Rolling in the Shadows: Analyzing the Extraction of MEV Across Layer-2 Rollups,” by Christof Ferreira Torres, Albin Mamuti, Ben Weintraub, Cristina Nita-Rotaru, and Shweta Shinde, quantifies activity and discusses novel MEV strategies on L2 that leverage the sequencer role and its L2 batch confirmation latency.

 

In addition to the above resources, Sorella Labs will soon release their MEV data indexer tool, Brontes, which will be an open source repository available for Ethereum mainnet and L2. Flashbots and the Uniswap Foundation are seeking grants to expand L2 MEV taxonomy and quantification. If you have worked in this area or are interested in collaborating, please contact the Flashbots market research team.

 

Although further validation is needed, the dataset published by hildobby on Dune Analytics provides a valuable initial reference standard.

 

Atomic arbitrage volume on L2 using the hildobby dataset

Data Sources

 

Over the past year, atomic arbitrage MEV trading volume on six major L2s, including Arbitrum, OP Mainnet, Base, Zora, Scroll, and zkSync, has exceeded $36 billion, accounting for 1% to 6% of all decentralized exchange (DEX) trading volume on each chain. These MEV trading volumes were initially concentrated on Arbitrum and OP Mainnet, but have recently gradually shifted to Base and zkSync.

 

Compared to atomic arbitrage volume, sandwich attack volume on L2 networks is significantly lower, in stark contrast to Ethereum, where sandwich attack volume is four times that of atomic arbitrage. This difference is mainly due to the single sequencer setup of the L2 network and the lack of a transaction pool, which limits the ability of searchers to perform sandwich MEVs using user transactions in the transaction pool unless there is a transaction pool data leak or a sandwich attack initiated by a single sequencer. Therefore, on L2, atomic arbitrage, blind backtracking, statistical arbitrage, and liquidation become more viable strategies for searchers.

 

Data Sources

 

Ethereum MEV volume breakdown

 

Measuring how much MEV revenue is left on L2 of the MEV market?

 

While it’s difficult to precisely quantify the MEV market, we can examine numbers from other ecosystems with MEV solutions for size comparison:

 

On Ethereum L1, annual validator revenue from MEV-boost blocks is ~$96.8M (based on an estimate of $3,500/ETH); the median value of a MEV-boost block is 4x the value of a normal validator block.

 

Block reward distribution for normal blocks and MEV-boost blocks

 

On Solana, the additional MEV revenue that validators collect from validator tips through Jito’s bundling service is estimated to be approximately $338M based on 50,000 SOL per week (based on an estimated price of $130/SOL).

 

Daily tips earned through Jito bundle service, by validator with Jito Labs

 

While the exact total MEV volume of the Base network has not yet been announced, we can estimate the market size by observing the revenue of the Banana Gun Telegram Bot, one of the most active players in the market. Banana Gun has roughly the same volume on Base’s L2 network as well as Solana, bringing in over $1 million in daily volume on each chain, which translates to over $10,000 in transaction fees per day on each chain.

 

Banana Gun Telegram Bot, cross-chain volume and costs

 

Please note that Banana Gun Bot's market share in Solana may be significantly different from Base. For example, several other major Telegram Bots exist on the Solana platform, such as Sol Trading Bot and BonkBot, while a smaller number of Telegram Bots may be supported on Base. Therefore, Banana Gun’s trading volume and MEV revenue ratio on Solana cannot be directly used to estimate the total MEV revenue on Base.

 

However, using another forecasting method, we can see different results: in March, the Banana Gun Telegram Bot paid more than $23 million to Ethereum block builders and validators. In particular, in the week of March 26-April 1, Banana Gun's transaction volume on Base actually exceeded that of Ethereum, as shown by the peak in the chart, which hints at the Base network's huge MEV revenue potential. This cross-chain transaction volume comparison reveals Base's growth prospects in terms of MEV.

 

Of course, there are significant differences in the MEV ecosystem between Base and Ethereum. Competition for MEV on Base may be less intense than on Ethereum, which may result in lower fees for bots to pay when bidding to validators. Despite this, those meme currency trading Bots that mainly rely on blind sniping and arbitrage mechanisms are still feasible under Base’s sequencer architecture.

 

Banana Gun Telegram Bot users’ MEV income paid to validators

 

Focus on MEV issues in L2 networks

 

Ethereum has developed a mature MEV ecosystem, equipped with infrastructure tools that serve participants at all levels of the supply chain. At the protocol level, MEV-boost allows validators to outsource block building tasks through auctions. For searchers, bundled services provided by Ethereum block builders - similar to Solana's Jito Labs and Polygon's FastLanes - enable them to implement MEV policies that include rollback protection. These services ensure that block builders simulate transactions and only execute those that are determined not to be rolled back. In addition, private RPC services like Flashbots Protect provide ordinary users with a way to bypass public transaction pools and their potential risks. However, the current L2 network still has a lot of room for improvement in developing MEV infrastructure comparable to this.

 

Why should we pay attention to MEV strategies and solutions for L2 networks?

 

The MEV phenomenon persists in an environment lacking trading pools and plays a key role in maintaining market efficiency, especially by executing strategies such as statistical arbitrage, atomic arbitrage, and liquidation to liquidate liquidity in outdated AMMs and lending markets.

 

However, the lack of mature MEV infrastructure, such as bundling services, may lead to some negative consequences. In the absence of a trading pool, many MEV strategies may degenerate into spam strategies, which will lead to:

 

· Increased network rollback rate;

 

· As a result, network congestion increases.

 

By implementing bundled services and shifting the focus of MEV competition from the main chain to the auxiliary chain, users can effectively reduce the burden of high gas fees due to MEV robot competition. At the same time, searchers can enjoy higher returns due to rollback protection, reducing the risk cost of failure.

 

For L2 networks that use a shared sequencer, current mainstream solutions often require users to publish transactions to a public transaction pool, which may lead to the recurrence of sandwich attacks. In this case, MEV protection tools like Flashbots Protect are particularly important. They can not only protect users from the threat of sandwich attacks, but may also provide refunds of MEV or priority fees, ensuring that users get better transaction execution and more favorable prices.

 

The development of complex MEV infrastructure faces several unresolved challenges. First, as more value flows to sequencers, searchers’ revenue patterns change over time, potentially reducing marginal profits. This change may raise questions about the sustainability of highly competitive search strategies in the long term. We expect market mechanisms to moderate this phenomenon, such that common search strategies will pay a larger portion, but not all, of their value to sequencers, while less common strategies will pay less.

 

Additionally, existing MEV infrastructure, such as Ethereum’s block construction market, is experiencing rapidly evolving order flow dynamics. To date, these factors have been the primary driver of the trend toward block construction market centralization and the rise of private trading pools on Ethereum L1. Ensuring that the block construction market remains competitive and fair remains an issue that needs to be addressed.

 

Finally, the MEV solution for L2 networks may need to be different from the current Ethereum mechanism, mainly due to the unique characteristics of L2: such as shorter block generation time, lower-cost block space, and relatively centralized governance structure. For example, Arbitrum has a block time of only 250 milliseconds, and it is not yet known whether such a fast block rate is compatible with the existing MEV infrastructure. At the same time, the ample and economical block space provided by L2 has greatly changed the transaction search landscape, making the spam problem more serious and urgently requiring new solutions. In addition, compared with other environments such as Ethereum L1, L2 has more centralized governance, which may allow additional requirements for MEV service providers, such as requiring block builders to avoid sandwich attacks on users to ensure market fairness.