TL;DR
This paper provides the first independent audit of Flashbots, a dominant "private transaction pool" designed to mitigate the negative effects of Maximal Extractable Value (MEV) on Ethereum. The researchers reveal that while Flashbots has successfully lowered gas prices for average users, it has also fueled extreme miner centralization and shifted the lion's share of MEV profits from independent "searchers" to a handful of powerful mining pools.
What Problem Does This Paper Address?
In the world of Decentralized Finance (DeFi), "Maximal Extractable Value" (MEV) is a polite term for the profit miners or savvy traders can make by reordering, including, or excluding transactions within a block. This often manifests as frontrunning (jumping the line to profit from someone else's trade) or sandwich attacks.
These activities create "negative externalities": they clog the network, spike gas fees for everyone, and can even threaten the stability of the blockchain itself. Flashbots was launched as a solution to move this "dark forest" of activity into a private, off-chain auction. The authors set out to determine if Flashbots is actually meeting its three core goals:
- Illumination: Making MEV activity transparent.
- Democratization: Allowing anyone to extract MEV, not just those with expensive infrastructure.
- Distribution of Benefits: Ensuring profits are shared fairly across the ecosystem.
Methods
The authors employed a data-driven measurement approach, analyzing Ethereum blockchain data from May 2020 to March 2022.
- MEV Detection: They used established heuristics to identify three types of MEV: sandwiching, arbitrage, and liquidations.
- Private vs. Public Identification: By comparing transactions seen in the public "mempool" (the waiting room for transactions) against those actually mined into blocks, they could infer which transactions were submitted through private channels like Flashbots or other "stealth" pools.
- Profit Analysis: They calculated the "insertion profit" for both miners and searchers by deducting transaction fees and "tips" (coinbase transfers) from the gross gains of an MEV attack.
Datasets
The study utilized several high-volume data sources:
- Ethereum Archive Node: Data from blocks 10,000,000 to 14,444,725.
- Pending Transactions: A corpus of over 125 million public pending transactions collected over five months.
- Flashbots API: Publicly available data on bundles and blocks mined through the Flashbots relay.
- Price Data: Token prices sourced via the CoinGecko API to normalize profits into Ether (ETH).
Key Insights and Ideas
- The Great Migration: Flashbots adoption was explosive. Within eight months of launch, it captured 99.9% of the Ethereum hashing power. It is no longer an "alternative" system; it is effectively a de facto part of the Ethereum protocol.
- The Miner's Windfall: Flashbots is a massive win for miners but a raw deal for "searchers" (the traders finding the MEV). Miners using Flashbots make 260% more profit than they did before, while searchers' profits have plummeted by over 84%.
- Centralization Paradox: Despite the goal of democratization, Flashbots blocks are highly centralized. Over 90% of Flashbots blocks come from just two mining pools (F2Pool and Flexpool).
- The "Other" Private Pools: Flashbots isn't the only game in town. About 13.2% of MEV extraction happens in other private pools or through miners acting solo, suggesting a fragmented and still-opaque landscape.
- Negative Profits: Surprisingly, about 1.58% of MEV attempts in Flashbots resulted in a net loss for the searcher, often due to faulty smart contracts or unexpected market shifts.
Key Results
- Total MEV: The researchers identified 4.5 million MEV extractions. While Flashbots accounts for about 31% of total extractions, it handles over 80% of the private MEV.
- Gas Price Relief: The paper confirms that Flashbots has successfully decoupled MEV competition from public gas prices, effectively lowering the "tax" on regular users.
- Profit Disparity: In sandwich attacks, miners' median profit rose from 0.048 ETH (pre-Flashbots) to 0.125 ETH (with Flashbots). Conversely, searcher profit dropped from 0.13 ETH to a measly 0.02 ETH.
Limitations and Future Work
The authors acknowledge that their heuristics provide a lower bound on MEV. They only looked at the three most common types of MEV and assumed sandwich attacks occur within a single block (which isn't always true).
Furthermore, the study was conducted before Ethereum's "Merge" to Proof-of-Stake. Future work is needed to see how the transition from miners to "validators" and the introduction of MEV-Boost has altered these dynamics. The authors also suggest that the current "sealed-bid auction" model of Flashbots inherently favors the auctioneer (the miner), and different auction designs might be needed to truly "democratize" the system.
Summary updated