Flashbots Research: How MEVs swallow the blockchain expansion dividend

転載元： chaincatcher

06/18/2025·5D

Author: Robert Miller, Flashbots

Compiled by: Saoirse, Foresight News

Today, we propose a new argument: MEV (maximum extractable value) has become the main limiting factor in blockchain expansion.

While mainstream public chains such as Ethereum, its Layer2 network, Solana, etc. are competing to expand at the fastest speed, the economic restrictions brought by MEV have emerged throughout the industry. On-chain search behavior is starting to occupy the main capacity of most high-throughput blockchains in an astonishing way of resource waste.

This is not a theoretical assumption or an individual phenomenon. This situation can be seen everywhere from Solana (MEV robots consume 40% of block space) to the Ethereum Layer2 ecosystem. To quantify impact, we conducted an in-depth analysis of the top OP-Stack Rollup that supports specific tracking endpoints, and the results reveal an industry-wide problem:

The junk trading robots in multiple Rollups consume more than 50% of Gas, but only pay less than 10% of the handling fee;
From November 2024 to February 2025, the Base network increased Gas processing capacity to 11 million Gas/second, but almost all of it was occupied by garbage robots (equivalent to the capacity of three Ethereum main networks!);
The ongoing demand for Gas from garbage robots has pushed up user handling fees;
The junk trading market is highly concentrated, with more than 80% of spam trading on Base dominated by two searchers.

Technical expansion methods such as database sharding (such as Rollup), validity proof, database or consensus mechanism optimization are important, but technology alone cannot solve the problem. Although we have mastered the methods of building basic technology throughput, the current market structure has imposed economic restrictions on expansion.

This article will analyze this market failure, demonstrate its impact with data, and propose a new MEV auction mechanism designed to solve the problem.

Analysis of spam transactions

To understand why block space is wasted, we first disassemble a successful arbitrage transaction:

Example of successful arbitrage transactions on Base

At first glance, this seems to be a model of efficiency: a search robot performs precision arbitrage, earns $0.12 and pays a $0.02 handling fee.

But the real cost of this successful arbitrage is shocking: for each successful arbitrage is completed, the robot sends about 350 transactions attempting arbitrage (mostly failing). On average, a single successful arbitrage consumes about 132 million Gas—equivalent to nearly 4 complete Ethereum blocks. It should be noted that this is just one of the many robots participating in the competition, and the actual cost of the chain is actually higher.

Now look at a typical failed attempt to understand the on-chain behavior of the robot:

Example of a failed transaction in blindly looking for arbitrage opportunities

At first glance, the transaction was not abnormal: the execution was successful and the token transfer was not carried out. The only clue is that it consumes about 2.6 million Gas (as shown in the picture above).

A deeper track of its internal calls reveals that it makes a series of calls to dozens of different DEX pools, querying the pool status through getReserves() and slot0(). These calls are essentially getting asset prices on different DEXs.

Shows a trace example of repeated calls for slot0 () and getReserves ()

The core logic of the robot is simple:

Send transactions to the chain
Query the price of multiple DEX pools during execution
If there is an arbitrage opportunity, execute it
If not, terminate the transaction

The above transaction is the embodiment of these four steps, and it is finally terminated and no operation is performed. In fact, it is just a high-intensity price query, consuming about 2.6 million Gas, but only reads the market status without substantial action.

On public chains such as Base, World and Solana, this strategy has become the mainstream way to extract MEVs. A few successful transactions need to pay for a large number of failed attempts, which is a rational choice for searchers, but it causes systematic inefficiency to the network.

A large number of resources are used to read prices without generating substantial value. And it is not the only one searcher, and all searchers have to adopt this strategy to capture atomic MEVs. The final result is as shown in the data: the public chain is blocked by spam transactions, and the handling fees are rising due to spam transactions. (Note: Atomic MEV emphasizes the value extraction achieved in single chain operations (such as a single transaction or within a single block), and is commonly found in scenarios such as arbitrage and rushing to utilize blockchain immediacy and transaction order.)

The root cause of spam trading

It is not accidental that high-throughput public chains are blocked by spam transactions, but a direct and "rational" reaction triggered by market structure defects: if a searcher wants to read the latest status of the block and make profits from it, he must blindly initiate transactions within the same block.

The arbitrage robot analyzed in the previous article is a typical case. Although off-chain query can obtain the status of the previous confirmation block, this lags behind the MEV opportunities that transactions are creating in the current building block. In networks such as Base or Solana, native memory pools (mempools) are private, which means that searchers cannot know the execution of user transactions and the opportunities they create before blocks are published. If you want to discover and capture the arbitrage space, the only way is to include your transactions in the same block immediately after the user transactions. Once you wait for the next block, the opportunity will be preempted.

The rampant on-chain search phenomenon stems from the interaction of the following factors:

1. Transaction expressiveness

Unlike traditional finance where traders submit simple static orders (such as "buy at X price"), searchers can create transactions as on-chain programs, embed conditional logic based on real-time state execution of the market to implement complex responsive strategies that were originally impossible.

2. Turn to private memory pools

To protect users from snatching, most high-throughput public chains set the memory pool to private. Although this can effectively defend against jumping, it also prevents searchers from seeing the user order flow. Since the transaction cannot react before it is on the chain, searchers can only blindly detect opportunities on the chain by initiating highly expressive transactions.

3. Low handling fees

The low-cost block space further amplifies on-chain search behavior. Searchers know that the profit of a single successful arbitrage can cover the cost of a large number of failed transactions, so they dare to send massive speculative transactions to each block. The lower the Gas fee, the more complex the searcher can write more complex logic and pursue more complex strategies. [1]

4. Lack of efficient auction mechanism

Competition among searchers lacks a formal mechanism for expressing transaction sorting preferences. Since it is impossible to bid directly for specific transactions in a block, competition degenerates into a wasteful alternative: consume more Gas. The main way for searchers to increase their winning rate is to consume Gas at more locations in the block to increase the probability that the transaction will fall in the "correct position".

These four factors have jointly spawned "spam trading auction", an extremely wasteful mechanism that not only contributes to network congestion, but also fails to effectively capture the value of MEV. To quantify the inefficient scale caused by spam transactions, we conducted data verification.

Research findings

Analysis confirms that MEV-driven spam transactions pose economic restrictions on capacity expansion.

We define spam transactions by identifying transactions that “repeat query DEX but do not transfer tokens”. This heuristic aims to position the systematic wasteful "backrunning" arbitrage that could have been completed off-chain but was forced to go on the chain. We implemented this method on both Python tools and Dune dashboards. For details on the specific methodology, please refer to the appendix.

Because spam detection tools rely on specific RPC methods, current data analysis is limited to OP-Stack Rollup. But data from the Ghost Logs team shows that Solana also has similar phenomena, and other Ethereum Rollups (such as ZKsync and Arbitrum) have also been found to have signs of spam trading.

1. Spam trading is systematic and universal

First, this problem is systematic and widespread. Analysis of OP-Stack Rollup shows that spam trading is not an isolated phenomenon, but a dominant force in the entire ecosystem. On chains such as Unichain, Base and OP mainnets, spam transactions usually consume more than 50% of total gas. It can be seen that this is a structural consequence of current market design, not a local anomaly.

2. The gas consumed by junk transactions far exceeds the handling fee

they pay

The second discovery shows that from the perspective of the chain, the efficiency of spam transactions is extremely inefficient.

In all the Rollups we analyzed, there is a huge gap between the resources consumed by spam and the revenue it generates. Compared with other users, the amount of gas consumed by the junk trading robot is several times the amount of payment fee. For example, garbage robots on the OP main network consume about 57% of Gas, but only pay about 9% of the handling fee, a 6-fold gap.

The gap between fee payment and Gas consumption shows that spam transactions bring huge external costs to the network without providing almost no corresponding value, which is a typical feature of a systematic inefficient market. This includes real waste of computing resources, as every full node is forced to execute these transactions, thereby increasing the hardware requirements of all network participants.

In addition, we analyzed how spam transactions in L2 affect Rollup's use of L1 data availability (Data Availability).

Data shows that in the one million blocks in February 2025, the garbage robots on Base contributed about 56% of Gas consumption, 26% of L1 DA (Data Availability Data Availability) usage, and 14% of on-chain fees. The proportion of DA usage of garbage robots initially surprised us, but then it was found that this was related to the proportion of transactions (rather than Gas consumption). This is reasonable because the amount of DA depends on the data compression efficiency, not the amount of Gas consumption.

3. Spam transactions limit and offset the benefits of expansion

Third, this inefficiency directly offsets the benefits of capacity expansion. To measure the negative impact of spam transactions, we introduced a new metric: effective Gas throughput, i.e., the amount of Gas available to users processed per second after Rollup deducts garbage robot consumption.

The trend for Base is particularly evident: In November 2024, the total Gas throughput was 15 million Gas/sec, while the effective Gas throughput of users was only 12 million Gas/sec. In the next four months, the total throughput increased by 11 million Gas/sec, but the effective throughput remained almost the same. In other words, almost all new processing power is occupied by spam transactions.

Interestingly, after the end of February, effective throughput began to be more consistent with the growth trend of total throughput. This seems to be related to market trading volume (and the resulting MEV): After the "Libra scandal" broke out on February 14, effective throughput began to grow again as the volume of Memecoin trading on Telegram robots fell.

4. Continuous demand for spam transactions pushes up user handling fees

Perhaps the most direct impact on users is that the continued existence of spam transactions artificially push up the baseline of transaction fees and keeps them high for a long time.

Although Rollup's expansion measures have reduced nominal fees to extremely low levels (for example, about US$0.01), making many natural users no longer sensitive to prices, theoretically, if the block space is sufficient, the user is not sensitive to prices, and the role of the EIP-1559 fee market mechanism, the fees should have approached the absolute minimum. The vision of expansion is to create enough capacity to make this nearly zero handling fee state the norm.

But this is not the case. Searchers trying to capture MEVs through spam transactions are filling blocks with massive transactions, consuming a lot of gas. This behavior pushes up block utilization, leading to a continuous increase in basic fees, which reflects more the systemic inefficiency of the MEV market than the real needs of natural users.

Although the handling fees borne by end users are still at a low level, the overall level is far higher than the actual requirements. The key point of this problem is that innovative application scenarios that rely on a large number of cheap block spaces (such as on-chain social networks or automated micropayments) are therefore excluded from the market.

5. The junk trading market is highly concentrated

Finally, the analysis shows that the searcher market for MEV spam trading is characterized by extreme centralization.

To verify this, we counted which smart contracts consume the most Gas classified as "spam" from block heights of 26000000 to 26900000. In preliminary observation, the market seems to have a high proportion of heads but its structure is scattered.

But this appearance is deceptive. On-chain analysis shows that the common strategy used by searchers is to rotate smart contracts used to send spam transactions, but transfer profits to a fixed "profit address". By tracking the ETH transfer path of successful arbitrage transactions, we try to identify smart contracts controlled by the same operator. Although not all robots adopt this mode, this is common for head robots.

When data are grouped by profit address, market concentration becomes extremely significant:

The result is obvious, and only two institutions dominate more than 80% of spam transactions on Base. This extreme centralization indicates that there are obvious barriers to entry in the market and that the current "spam auction" is not a truly competitive market. The lack of competition further weakens the price discovery mechanism, resulting in public chains being unable to capture the true value of extracted MEVs and having to bear the negative externalities brought about by spam transactions.

The way forward

We believe that blockchain should maximize valuable economic activity in a limited block space.

From this standard, the current "spam trading auction" mechanism is extremely inefficient: it only takes about 200,000 Gas to complete two redemptions on Uniswap v3, while it takes about 130,000 Gas to achieve the same economic results on Base. The efficiency gap is as high as 650 times, and narrowing this gap is the key to unlocking the real potential of expansion.

To solve this problem, we must first go back to the four major reasons why on-chain search has become the mainstream model: transaction expression, memory pool privacy, low handling fees and lack of efficient auction mechanisms. Among them, low Gas fees and high expression are the clear goals of the general smart contract chain [2], and we need to continue to strengthen these characteristics. Therefore, the solution must focus on two other points: allowing searchers to read the state of the upcoming link and express their preferences in a way that not only protects users' rights and minimizes spam transactions on the chain.

Solution Direction

1. Realize state transparency through programmable privacy

An efficient market requires the searcher to provide real-time access to transaction flows, while programmatically limiting the way they use information. The system needs to verify that searchers can only conduct "backrun" transactions and cannot implement frontrun, sandwich attacks, or leak privacy data. This visibility allows searchers to perform conditional logic off-chain instead of blindly detecting on-chain. When a searcher generates potential profitable transactions off-chain, it still needs a way to accurately embed it into the block to capture the MEV.

2. Build an MEV auction mechanism for explicit bidding

Abandon the "spam trading auction" model with Gas consumption as the competitive dimension, and instead design a transaction sorting rights bidding mechanism based on economic incentives. Searchers can directly submit currency quotes for the block location of the target transaction, and determine the transaction sequence through a market-oriented pricing mechanism. This model converts disordered Gas consumption competition into an efficient price discovery process:

Searchers do not need to send hundreds of invalid transactions, they only need to pay for truly valuable sorting rights;
Blockchain can capture the true value of MEVs through auctions, rather than waste resources on meaningless chain computing.

Flashbots are already trying to leverage trusted execution environments (TEEs) to provide visibility to searchers while preventing sandwich attacks. TEEs ensure that the data remains confidential even to the machine operator when a specific code is executed.

This allows searchers to run in TEE, verbally run back on private transactions without the ability to implement sandwich attacks or export any privacy data. We have verified this model on Ethereum L1, and searchers have been conducting post-run transactions through similar systems for months and are actively adapting it to L2.

in conclusion

Discussions on scaling have long been limited to basic technology throughput. However, our research shows that the key breakthrough point is no longer to expand block capacity, but to use block space more efficiently [3]. This is because every unit of block space is released, the MEV incentivizes spam transactions to consume new capacity. In other words, most of the benefits brought by "expansion" are captured by economically rational MEV robots, and real users cannot benefit from it. This problem is pushing up the handling fees of ordinary users, restricting the effectiveness of capacity expansion, and causing a large amount of network resources to be wasted.

The limitation of capacity expansion lies in this: although increasing block space can increase throughput, the improvement of handling fees is limited, because increasingly complex on-chain MEVs will swallow most of the gain. To break through these restrictions and unlock the real potential of expansion, we must get rid of the wasted junk trading market. Through programmable privacy and explicit bidding, we can eliminate the incentives for spam trading and replace "spam trading auctions" with a expressive, fair and efficient MEV market.

Adopting MEV auctions is not a luxury option, but a strategic need. The core is to use TEEs to provide searchers with access to transaction flows, while also restricting their usage by programming. This design achieves ideal results: support back-run arbitrage without spam trading while preventing sandwich attacks. For blockchain, this means capturing more revenue in an efficient, garbage-free market; for users and developers, lower and stable handling fees and real available capacity will ultimately unlock the full value of the expansion.

What will happen to the world when we break through the limits of spam trading? When the transaction costs are so low that they are almost negligible, which new ones may be unlocked? What new applications will be born? The answer can only be proved by practice.

Thanks to DataAlways, Hasu, Fahim, Danning, dmarz, Nathan, Georgios, Dan, buffalu, Quintus, Tesa, Anika, Brian, Xin, Sam, Eli, Christine, Christoph, Alex, Fred, and many others for their valuable comments. Special thanks to Phil, and also to Achal for his help in design.

appendix

Spam transaction identification heuristics

To identify spam transactions, we adopt two heuristic rules:

Tokenless transfer: Does the transaction involve any token transfer? If so, it is not classified as spam.
Repeat DEX price query: If a transaction initiates at least 4 queries on common DEX price data without performing token transfers, it is classified as a spam transaction.

We believe that at the time of writing these heuristics are reliable: Any operation involving token transfers usually has real value to the user, and spam transfers tokens only when MEV opportunities are captured. In addition, DEX price query rules can effectively identify robots that systematically detect arbitrage opportunities, which is the main form of spam trading we have observed. This definition focuses on wasteful behaviors that only query DEX prices on the chain, excluding productive backrunning behavior.

However, this definition needs to be further optimized in the future: spam trading robots can bypass this rule by simply transferring tokens, so the classification criteria for "spam trading" are still worthy of subsequent research. In addition, this definition mainly covers blind post-run arbitrage robots that dominate MEV mainstream, and does not include other MEV strategies such as liquidation.

Methodology of junk transaction identification

We identify spam transactions by analyzing transaction tracking: for each transaction, check all its tracking to determine whether to call the token transfer function or DEX price function (such as slot0(), getReserves(), etc.). If the transaction involves token transfer, it is excluded; if the token is not transferred and 4 or more DEX price inquiries are initiated, it is classified as spam transaction.

The choice of 4 times as the threshold was conservative, and the experiment showed that setting the threshold to 3 times had little effect on the overall result. Similarly, we filter transactions by transfer events on Dune and found that it differs significantly from the tracking-based method results.

spam-inspect tool

To study spam-inspect, a Python tool designed to analyze Ethereum Rollup activities, designed to efficiently identify spam-inspect. The tool enables analysis by tracking every transaction within a block and using the heuristic rules mentioned above.

This tool relies on the trace_block method and is currently only available on OP-Stack chains that support OP-Reth or OP-Erigon.

Dune Query

We built materialized views on Dune, positioning hashes that meet spam trading standards by filtering transactions containing Transfer events and identifying duplicate DEX price calls. The difference from spam-inspect is that this method relies on transfer events rather than transaction tracking. These spam transaction materialized views are used for subsequent query analysis.

Data Availability (DA) Estimation

Although this article mainly discusses the impact of spam on Gas, it can also consume other resources, such as the use of Rollup on L1 data availability. To estimate the L1 DA resources wasted by L2 spam transactions, we built a custom data pipeline (some modules that reuse op-batcher) and obtained the results through two sets of calculations: