Flash Boys 2.0: Frontrunning in Decentralized Exchanges, Miner Extractable Value, and Consensus Instability. Published in 2020 at the IEEE Symposium on Security and Privacy, this work by Philip Daian and colleagues follows bots competing to place profitable transactions ahead of one another. The security consequence for exchange operators and consensus-security teams is that miners may profit through transaction inclusion, exclusion and ordering, not just visible fees. The researchers measured that competition and examined what it implies for consensus incentives, but their findings do not prove that an extractable-value incentive caused a consensus reorganization.

The study isolates what it calls pure revenue: an atomic sequence of trades that produces net revenue in every asset involved. That excludes strategies which finish while still holding an asset whose price could move against the trader. Bots chased these opportunities through priority gas auctions, repeatedly replacing transactions with versions offering higher fees to win favorable ordering. MEV, which stands for Miner extractable value, covers the total ether a miner can extract by manipulating transactions over a chosen period, potentially across multiple blocks. It includes value from deciding which transactions appear and in what order.

Taken together, the measurements and modeling address two linked questions. What does automated competition for exchange ordering look like, and how much value can be observed? Then, if miners can include, exclude or reorder transactions to capture that value, could the resulting incentive threaten consensus stability? The researchers examined visible competition and feasible reordering strategies. They did not establish that MEV had actually caused a consensus reorganization.

The researchers operated geographically distributed, modified Ethereum nodes running geth. They synchronized the nodes’ clocks and collected more than 300 gigabytes of network observations. They supplemented these pending-transaction observations with on-chain records from Google BigQuery and price data from CoinMetrics. They also analyzed a deliberately simplified model of competition in priority gas auctions and released associated code and public data.

The measured market was valued in the millions of US dollars at the time of the experiment, but that amount was explicitly a lower bound rather than a complete accounting of extractable value. For the observed pure-revenue transactions, the median realized profit was 65 percent of the opportunity the researchers identified. In plain terms, the transaction in the middle of that distribution retained about two-thirds of the available opportunity as profit.

One captured block contained a 101.6-ether arbitrage opportunity alongside the conventional block reward and 0.022 ether in explicit fees. In that observed case, the value available through transaction ordering greatly exceeded both conventional payments combined. It is a concrete example of the incentive the study examines, not evidence that every block contained a similarly large opportunity.

The auction model intentionally reduces competition to 2 players. It also assumes fixed latency and rate limits, so real competitors need not behave the same way. Detection covered only selected popular exchanges and suspected bots. The method could merge unrelated auctions or miss low bids, losing transactions and unrecognized bot addresses. Profit calculations omitted infrastructure costs and many failed transactions. The analysis also did not fully group addresses controlled by the same actor. Finally, the researchers demonstrated incentives and feasible strategies, not an MEV-driven consensus reorganization in the observed network.

For detection engineers, repeated transaction replacements with rising fees can indicate competition for ordering. Protocol and security architects should include the value available through inclusion, exclusion and ordering in their threat models instead of considering explicit fees alone. Financial estimates based on this study should remain labeled as lower bounds, and operational profit calculations should account for infrastructure and failed transactions. These are defensive interpretations of the evidence; they do not turn the demonstrated incentive into proof that a consensus attack occurred.

Daian and colleagues combined network measurement of priority gas auctions with a broader account of value miners can obtain by controlling transaction inclusion and order. Detection teams, decentralized-exchange operators and consensus designers should ask who controls ordering and what rewards that control creates. They should not treat the measured amount as a complete estimate of all extraction or cite this work as evidence of an observed consensus reorganization. Its supported takeaway is narrower: profitable ordering competition was measurable, and the resulting incentives deserve explicit security analysis.