MEV Strategies in 2026: The Verdict and the Dark Forest
A US judge just cleared the brothers behind a $25 million MEV exploit, ruling that Ethereum has no terms of use. Here is what the bots that reorder your trades actually do.
On 16 November 2026, a federal judge in Manhattan did something no crypto trader had seen before. She looked at the most-watched MEV prosecution in history, a case the government had billed as the first criminal case over Ethereum trading bots, and ruled that there was no crime to answer for. US District Judge Jessica Clarke acquitted Anton and James Peraire-Bueno, the MIT-educated brothers accused of draining roughly $25 million from Ethereum trading bots in about twelve seconds. Her reasoning reduced to a single sentence that DeFi lawyers will be quoting for years: Ethereum, she wrote, is a trustless, highly adversarial environment with no terms of use.
That ruling did not just end one case. It put a legal spotlight on an entire hidden economy: the business of Maximal Extractable Value, or MEV, the profit that bots and professional firms extract by deciding the order of transactions inside each block. MEV is the invisible tax on nearly every on-chain trade, the reason your swap sometimes fills at a worse price than the screen promised, and a market that has moved more than a billion dollars most users never see. This is a guide to the strategies that drive it, the supply chain that sells it, the defenses that blunt it, and what a judge’s phrase, no terms of use, actually means for the people building and fighting it.
What MEV Actually Is
Maximal Extractable Value is the most value that can be squeezed out of producing a block, over and above the standard block reward and the gas fees users pay, by including, excluding, or reordering the transactions inside it. The term was coined in the 2019 research paper Flash Boys 2.0 by Phil Daian and co-authors, who warned that the race to capture it could destabilize consensus itself. It was originally called Miner Extractable Value; after the Merge moved Ethereum to proof-of-stake, where validators rather than miners order blocks, the industry kept the acronym and relabeled it Maximal.
The mechanics are easier to see with a number. Suppose a decentralized exchange is quoting ETH slightly cheaper than the rest of the market. A bot can buy on that venue and sell where the price is higher inside a single transaction; Ethereum’s own documentation uses the example of a 1,000 ETH trade that comes back as 1,045 ETH, a 45 ETH profit that exists purely because someone acted first. Multiply that across every block, every pool, and every price gap, and you get an industry. Cumulative MEV extracted on Ethereum has run past $1.2 billion, with sandwich attacks alone accounting for more than half of it.
The culture around MEV has its own vocabulary. In August 2020, Paradigm researchers Dan Robinson and Georgios Konstantopoulos described the public mempool, the waiting room where pending transactions sit before they are mined, as a dark forest: a place where any transaction worth money is prey, and where sending a naive trade into the open is like lighting a fire in the woods at night. Five years later, that metaphor is still the single best way to understand why on-chain trading feels the way it does.
The Verdict That Said the Dark Forest Has No Rules
The Peraire-Bueno case is worth understanding in detail, because it is the clearest test yet of whether MEV can be a crime. In April 2023, the brothers ran their own Ethereum validator. They sent eight carefully crafted bait transactions into the public mempool, designed to look like juicy prey and lure the sandwich bots that hunt there. When the bots took the bait and built their attacking trades around it, the brothers sprang a trap on the plumbing itself.
The flaw they exploited lived in the MEV-Boost relay, the piece of infrastructure that hands finished blocks to validators. The relay was supposed to reveal a block’s contents only after the validator had committed to it, but it released the transactions before fully validating the signed block header. The brothers’ validator proposed a block with two falsified header fields; once the relay revealed the bots’ transactions, they reordered the block to drain the attackers instead, walking away with about $25 million in roughly twelve seconds. Flashbots documented the underlying relay weakness in its own April 2023 post-mortem, the same incident at the center of the trial.
The Department of Justice indicted the brothers in May 2024 on three counts: wire fraud, conspiracy to commit wire fraud, and conspiracy to commit money laundering. The first trial ended in a mistrial in November 2025 when the jury deadlocked after three days. Rather than retry, prosecutors watched Judge Clarke enter a judgment of acquittal on all counts. Her logic was narrow and devastating: wire fraud requires a material misrepresentation, a lie that matters to the party being deceived, and she found none. The bait transactions promised nothing, because no trader who posts a transaction to the public mempool would ever agree to be sandwiched. The falsified signature, she noted, had no effect on the behavior of its recipient, because the relay never checked the fields in the first place. Conduct that is improper, objectionable, or even despicable, she wrote, is not by itself wire fraud.
A Field Guide to MEV Strategies
MEV is not one thing. It is a spectrum of strategies that run from genuinely useful to openly predatory, and lumping them together is the single most common mistake in the debate. Some keep markets efficient and lending protocols solvent. Others are pure extraction from an unsuspecting counterparty. The table below is the map most professionals carry in their heads.
| Strategy | How it works | Who pays | Harm level |
|---|---|---|---|
| Arbitrage | Buy low on one venue, sell high on another, aligning prices | Diffuse; the counterparties who mispriced | Benign, often useful |
| Liquidations | Repay an unhealthy loan for a bonus when collateral drops | The over-leveraged borrower | Benign, keeps lending solvent |
| Sandwich | Buy ahead of a victim swap, let it fill worse, then sell | The retail swapper | Toxic |
| JIT liquidity | Add liquidity for one block around a big trade, then pull it | Passive liquidity providers | Gray |
| Backrunning | Trade immediately after a price-moving transaction | The broader market | Mostly benign |
| Long-tail and OEV | Capture value from oracle updates and niche events | Protocols and their users | Gray |
| Cross-domain | Extract value spanning chains, rollups, or CEX and DEX | Cross-venue traders | Gray to toxic |
Arbitrage and Liquidations: The Benign End
Start with the strategies almost nobody objects to. Arbitrage is the oldest trade in finance, and on-chain it does something genuinely valuable: it drags the price on a decentralized exchange back into line with everyone else. Every swap that moves a price on an automated market maker opens a small gap against the wider market, and the same curve math that prices every trade on an AMM is what tells an arbitrage bot exactly how large that gap is and how much capital will close it. Without these bots, DEX prices would drift and stay wrong.
The largest and least visible form is CEX-DEX arbitrage, which bridges the price on a centralized exchange to the price on-chain. It is non-atomic, meaning one leg executes off-chain where nobody can see it, and it dwarfs the flashier on-chain strategies. A 2025 study by Fei Wu, Danning Sui, Thomas Thiery, and Mallesh Pai measured roughly $233.8 million in CEX-DEX extracted value from more than seven million arbitrages run by just nineteen searchers over nineteen months, with the top three capturing around three-quarters of the value. This is the darkest corner of the dark forest: the most profitable MEV strategy is also the one retail users will never watch happen.
Liquidations are the other respectable strategy. When a borrower’s collateral in a lending market falls below the required threshold, a liquidator steps in, repays part of the loan, and claims the collateral at a discount. It looks brutal to the borrower who gets liquidated, but it is the mechanism that keeps a protocol from accruing bad debt, and the competition to perform it quickly is a textbook MEV race. Most of the time, the fastest bot wins and everyone else’s deposits stay safe.
The Sandwich: MEV’s Most Hated Strategy
Then there is the sandwich, the strategy that gives MEV its bad name. The mechanic is simple and merciless. A bot spots your pending swap in the mempool. It buys the same asset immediately ahead of you, pushing the price up. Your trade then executes at that worse price. The bot sells right after, pocketing the bump. You paid the difference, and it shows up on your statement as slippage you assumed was just market movement.
No bot embodies this better than the one trading under the name jaredfromsubway.eth. In April 2026 it sandwiched a four-dollar token swap by Vitalik Buterin himself, wrapping more than a million dollars of volume around a trade worth pocket change; the bot barely profited, but the symbolism was perfect. Weeks later, in a twist the whole industry enjoyed, the same bot was drained of about $7.5 million in an exploit of its own contracts. The predator became prey, which is the dark forest working exactly as advertised.
The sandwich is also where the acquittal bites. Judge Clarke’s observation that no trader would agree to be sandwiched was not a defense of the practice; it was a finding that the victims of a sandwich are not being lied to, because nobody in the mempool consents to anything. That matters enormously for the law, because it means the millions of sandwiches that happen every year are, on the current reading, not fraud, even when the thing being sandwiched is itself a predatory bot.
The Gray Middle: JIT, Backrunning, and Oracle Value
Between the useful arbitrage and the hated sandwich sits a gray middle that most users never think about, and it is where a lot of recent MEV innovation actually lives. Just-in-time liquidity, or JIT, is the cleverest example. A sophisticated liquidity provider watches for a large incoming swap, adds a huge amount of liquidity to the pool in the same block to capture most of that swap’s fee, then withdraws it immediately after. The trader often gets a slightly better price, but the passive liquidity providers who keep their capital in the pool day and night watch the fee they expected get skimmed by someone who showed up for a single block. It is not theft, exactly; it is the professionalization of a role that used to reward patience.
Backrunning is gentler. Instead of jumping ahead of your trade, a backrunner places its own transaction immediately after a price-moving event, capturing the arbitrage your trade created without degrading your fill. Because it does not worsen the victim’s price, backrunning is the strategy most order-flow auctions try to harness rather than block, turning what would have been a pure loss into a rebate. The newest frontier is oracle-extractable value, or OEV: the profit available in the instant a price oracle updates, when fresh liquidations and arbitrage suddenly become possible. A growing set of protocols now auction the right to act on their own oracle updates, trying to recapture value that used to leak straight to searchers. The through-line across all three is the same: value that the naive version of DeFi left on the table is being systematically hunted, priced, and, increasingly, redistributed.
The Supply Chain: Searchers, Builders, Relays, Proposers
MEV is not extracted by a lone genius at a keyboard. It flows through an industrial assembly line that Ethereum formalized under the name Proposer-Builder Separation, or PBS. Understanding that chain is the key to understanding both the Peraire-Bueno exploit and the fixes now being built to prevent the next one.
At the front are searchers, who scan the chain for opportunities, bundle the transactions that capture them, and bid for the right to have those bundles included. Builders collect bundles and public transactions and assemble complete, ordered blocks, competing in a sealed-bid auction to produce the most valuable one. Relays sit in the middle as trusted couriers: they hold a finished block, show the validator only a header, and reveal the contents after the validator has signed. Proposers, the validators chosen to propose a block, simply sign the most valuable header offered to them. More than ninety percent of Ethereum blocks are now produced through MEV-Boost, the software that runs this auction.
| Role | What they do | How they are paid |
|---|---|---|
| Searcher | Find MEV, build bundles, bid for inclusion | The spread left after the bid |
| Builder | Assemble the most valuable full block | A cut of block value, plus tips |
| Relay | Hold the block, reveal only after the signature | Usually unpaid; a trust service |
| Proposer (validator) | Sign the winning header blind | The winning bid, added to staking yield |
The structural weakness is right there in the word blind. The proposer signs a header for a block it cannot see, trusting the relay to be honest about the contents. That trust is precisely what the Peraire-Bueno brothers broke in 2023: they were the proposer, and they found a moment where the relay revealed the contents before the signature was truly binding. Every defense discussed later in this piece is, in some sense, an attempt to remove that moment of blind trust.
The Builder Oligopoly
The promise of PBS was decentralization: anyone could build blocks. The reality in late 2026 is an oligopoly. On the day this piece was written, data from relayscan.io showed a single builder, Titan, producing about 57 percent of all MEV-Boost blocks over the previous day, with Quasar at roughly 19 percent. The top three builders together accounted for around 84 percent of blocks. A handful of relays, led by ultrasound.money and Titan’s own relay, route almost everything.
Concentration this tight is not an accident; it is the product of vertical integration. Titan launched its own relay in 2024 that accepts only its own blocks, folding two links of the supply chain into one firm. Searcher-builders that both find MEV and build blocks can keep value that independent searchers have to share. The main counterweight is BuilderNet, a project from Flashbots, Beaverbuild, and Nethermind launched in late 2024 that builds blocks inside trusted execution environments and shares the resulting MEV and fees with the users whose order flow created it. That same confidential-computing hardware is quietly becoming load-bearing across crypto, from block building to the market where verifiable compute finally found a paying customer in 2026.
Concentration also enables censorship. After the US sanctioned the Tornado Cash mixer in 2022, several relays began filtering transactions involving sanctioned addresses, and the share of blocks built by censoring infrastructure became a live governance fight. The tracker at mevwatch.info exists specifically to show how much of the network is willing to exclude a legal-but-sanctioned transaction, a reminder that whoever controls block ordering controls more than just profit.
MEV as an Invisible Tax
The cleanest way to think about MEV’s cost is to borrow a term from traditional markets: payment for order flow. In US equities, retail brokers sell their customers’ orders to market makers who profit from the spread. MEV is the on-chain version of the same bargain, except nobody signed a contract and the payment is extracted at the moment of execution. The value in your order flow is real, and in the dark forest it belongs to whoever sees your transaction first.
Competition has pushed searcher margins toward zero, which sounds like good news until you notice where the value went: up the chain, to builders and validators, as higher bids and tips. The visible swap fee a DEX charges is only a slice of what a trade really costs once the implicit MEV is counted; one academic analysis of the all-in cost of DEX trading estimates the fee you can see at only about a third of the total, with the rest hidden in slippage and extraction. And the tax is not constant. It concentrates when markets are stressed and liquidity is thin, which is exactly the condition crypto is in as this is published, with ETH trading around $1,659 and down roughly 8.6 percent on the week. Thin books mean wider gaps, and wider gaps mean more extractable value flowing to fewer, larger players. The macro backdrop, including the Fed decision looming over December, is not separate from the MEV story; it is the weather in the forest.
How to Defend Yourself: The Three Layers
The good news is that an ordinary user is not helpless. Three layers of defense exist today, and each one can be switched on in minutes. The first is the simplest: keep your transaction out of the public mempool entirely. Private RPC endpoints such as Flashbots Protect and MEV Blocker send your trade straight to builders without broadcasting it to the dark forest, so the bots never see the prey. MEV Blocker goes further and runs an order-flow auction on any backrunning opportunity your trade creates, returning the large majority of that value to you rather than to a searcher.
The second layer is those order-flow auctions themselves, exemplified by Flashbots’ MEV-Share, which lets users opt to reveal just enough about a transaction to capture a backrun bid and returns around ninety percent of the resulting value. The third and most structural layer changes the trade model altogether. Intent-based systems do not broadcast a specific transaction at all; they broadcast a desired outcome and let professional solvers compete to fill it. UniswapX runs a Dutch auction and can fill from a solver’s own inventory, CoW Swap batches orders and settles them at a single uniform clearing price, and 1inch Fusion blends request-for-quote with a falling price. Because there is no single victim transaction sitting in the open, the sandwich simply has nothing to bite.
| Defense layer | Example | What it stops |
|---|---|---|
| Private order flow | Flashbots Protect, MEV Blocker | Front-running and sandwiches (bots never see the trade) |
| Order-flow auctions | MEV-Share, MEV Blocker rebates | Returns backrun value to the user |
| Intents and batches | UniswapX, CoW Swap, 1inch Fusion | Removes the single victim transaction entirely |
None of this is a substitute for basic hygiene. Set a tight slippage tolerance so a sandwich has no room to work, use fresh deadlines so a stale transaction cannot be held and replayed, and choose the venue that routes through protection by default. The same instinct that keeps you safe from a well-crafted phishing attack, assume the open network is adversarial and verify before you sign, is exactly the instinct that keeps you from being sandwiched.
Solana’s Different Dark Forest
MEV is not an Ethereum-only phenomenon, but it looks different on Solana. Solana has no global public mempool; transactions are forwarded directly to the current block leader, which removes the open waiting room that makes Ethereum front-running so easy. In its place sits Jito, whose client runs on the large majority of Solana stake and lets searchers submit bundles with tips attached. The JitoSOL liquid staking token recycles those tips back to stakers, turning MEV into a visible component of yield rather than a hidden tax.
Solana has also been more willing to police the worst behavior socially. In 2024 the Solana Foundation removed more than thirty validators from its delegation program for running private mempools that enabled sandwich attacks on retail users. Tim Garcia, the foundation’s validator relations lead, described the purge as final, a stance that stands in sharp contrast to Ethereum’s more neutral, let-the-market-sort-it-out culture. Neither approach has eliminated MEV; they have only chosen different places to draw the line.
The Law After the Acquittal
It would be easy to read the Peraire-Bueno acquittal as a declaration that MEV is legal. It is not that, and the details matter. Clarke’s ruling is a district-court decision, fact-specific and not binding precedent, and several observers have noted that it leaves the broader law of on-chain conduct genuinely unsettled. What it did was reject one specific theory of criminality, cleanly and on the merits.
That theory had a name: the honest validator theory, the government’s argument that a validator owes some duty of honest dealing that the brothers violated. The defense answered with a version of code is law, that the only rules on a permissionless chain are the ones the software enforces. Coin Center filed an amicus brief backing the defense, and its argument is worth stating plainly. Honest validation, the group wrote, simply means obeying the specified rules of consensus articulated in the protocol software; importing a separate, prosecutor-made code of honest dealing would replace verifiable protocol enforcement with after-the-fact moral judgment. Coin Center’s Peter Van Valkenburgh put the chilling effect bluntly: anyone participating in an open network would have to wonder whether prosecutors might later declare their behavior dishonest and haul them into court.
The acquittal did not come out of nowhere. It rhymes with the Mango Markets case, where Avraham Eisenberg drained roughly $110 million in October 2022, was convicted in April 2024, and then had his convictions vacated in May 2025 on strikingly similar logic: a permissionless protocol with no rules prohibiting the conduct cannot be the site of a material misrepresentation. The government’s appeal in that case is still pending before the Second Circuit, which means the single most important open question in MEV law, whether an appellate court will agree, has not yet been answered.
There is a live counter-argument, and it is not fringe. Legal scholars including Mikolaj Barczentewicz have argued in work for the International Center for Law and Economics that blockchain transaction ordering can constitute market manipulation under existing doctrine. And the regulators have their own tools: the SEC’s Rule 10b-5 and the CFTC’s Rule 180.1 both reach manipulation on markets they oversee. The perimeter gap is the real issue. An anonymous bot sandwiching a stranger on a permissionless DEX sits outside those rules in a way a registered broker front-running a client never could, and that gap looks a lot like the broader enforcement blind spots crypto keeps rediscovering. The lesson for a trader is simple: do not count on the courts to protect you from MEV.
The Protocol’s Own Answer: ePBS and Beyond
If the courts will not referee the dark forest, maybe the protocol can. The next major Ethereum upgrade, Glamsterdam, makes enshrined Proposer-Builder Separation, EIP-7732, its headline feature. The idea is to fold the builder auction into Ethereum’s own consensus rules and remove the trusted relay, the exact piece of infrastructure the Peraire-Bueno brothers exploited. Done right, the blind-signing moment disappears, and with it a whole category of relay-trust failures.
Shipping it is proving hard. Glamsterdam ran on the Sepolia testnet on 6 October 2026, but developers called off the planned Hoodi testnet fork scheduled for 26 October, and as of the Ethereum Foundation’s early-November update there is no new Hoodi date and no mainnet date. Even when it lands, it will not be a cure. Vitalik Buterin has argued that ePBS alone is not enough to curb builder centralization, and that Ethereum will also need tools like FOCIL, which uses a random committee to force transaction inclusion and resist censorship, encrypted mempools that hide transaction contents until ordering is fixed, and anonymized routing. That is a multi-year research agenda, not a patch.
What It Means for You
Put the three threads together and the picture is clear. The courts have now said, at least once and at least at the trial-court level, that they will not convict someone for playing the dark forest by its own rules. The protocol’s fix for the worst structural weakness is real but unscheduled, still stuck on a testnet with no date for the mainnet. And the economics keep pulling block production toward a handful of dominant builders, especially when markets are thin and stressed.
What is left is the part you control. The defenses that work today, private RPCs, order-flow auctions, and intent-based trading, do not require a verdict or a hard fork. They require you to stop broadcasting naked transactions into a forest full of predators. MEV is not going away; it is becoming more professional, more concentrated, and, after 16 November 2026, more clearly beyond the reach of a fraud statute. The reorder tax on your trades is now a permanent feature of on-chain life. The only real question is whether you pay it by default or opt out.
Frequently Asked Questions
What is MEV in crypto?
MEV, or Maximal Extractable Value, is the profit a block producer or bot can capture by including, excluding, or reordering transactions in a block, beyond normal fees. It is the main reason an on-chain swap can fill at a worse price than you expected.
Were the Peraire-Bueno brothers convicted of MEV fraud?
No. Their first trial ended in a mistrial in November 2025, and on 16 November 2026 US District Judge Jessica Clarke acquitted both brothers on all counts, ruling that Ethereum has no terms of use and that no material misrepresentation occurred.
How do I protect myself from sandwich attacks?
Route your trades through a private RPC such as Flashbots Protect or MEV Blocker so bots never see them, use intent-based exchanges like UniswapX or CoW Swap, and set a tight slippage tolerance so an attacker has no room to work.
Is MEV illegal?
Most MEV, including arbitrage and liquidations, is legal and useful plumbing. The 2026 acquittal suggests that even aggressive strategies on permissionless chains are hard to prosecute as fraud, but the law is unsettled and the SEC and CFTC still police manipulation on regulated markets.
What is ePBS and will it stop MEV?
ePBS (EIP-7732) is an Ethereum upgrade that puts proposer-builder separation into the protocol and removes the trusted relay. It will redistribute and help decentralize MEV rather than eliminate it, and as of November 2026 it has no mainnet activation date.
Yuki Tanaka covers DeFi, market structure, and on-chain infrastructure for HOGE Wire.