You swap 5 ETH for USDC. The interface quotes a price.... You get a worse one.... Notihng broke, nothing got hacked, the transaction confirmed exactly as it was supposed to.... Somewhere between the moment you signed and the moment the block landed, software read your trade, positioned itself around it, and took a slice.
That slice has a name. maximal Extractable Value, or MEV, is the profit that whoever decides the order of transactions in a block can pull out purely by choosing that order. Not by hacking anythin. Not by breaking any rule. Ordering is a lever, and someone is always holding it.
Most explanations of MEV get lost in validator economics and builder auctions within two paragraphs. , This one starts where it actually touches you, which is the price you got versus the price you were shown.
What is MEV in crypto and where does the money come from
Every blockhain has a moment between transaction broadcast and transaction final where the contents are known but the position is not yet fixed. On Ethereum that gap lives in the public mempool..... your pending swap sits there, fully readable, for a few hundred milliseconds to a few seconds... And anyone running a bot can see exactly what you are about to do and what it will do to the price. More importantly, The vlue that can be squeezed out of that window breaks down into a handful of recognisable strategies.. Trackers disagree on absolute dollar totals because they classify transactions differently and cover different date ranges, so treat the totals you see quoted anywhere with suspicion. The proportions are far more stable than the headline numbers.| MEV type | Roughly what share | Who pays for it | Is it harmful to you |
|---|---|---|---|
| DEX arbitrage | ~35% | Liquidity providers on the stale pool | Mostly no, it aligns prices acros venues |
| Sandwich attacks | ~30% | The person doing the swap | Yes, directly and entirely |
| Liquidations | ~25% | The borrower being liquidated | Neutral, the position was going to close anyway |
| Everything else (NFT sniping, long tail) | ~10% | Varies | Situational |
That table is the single most useful thing to understad about MEV..... Only one of those four rows is money taken out of a normal users pocket.... Arbitrage is genuinely useful and keeps prices honest across venues. Liquidations are the system working as designed...... Sandwiching is the one that is pure extraction, and it is close to a third of the total.
How a sandwich attack actually takes money out of your trade
The mechanics are simpel enough to explain in four steps, which is exactly why it is so profitable.
Step one. You submit a swap of 5 ETH into USDC on an automated market maker, with slippage tolerance set to the default. Most interfaces default to somewhere between 0.5% and 1%. That number is not a safety setting. It is a permission slip. Sure, Step two... A bot reads your pending transaction and simulates it. It calculates exactly how far your trade will move the pool price and exactly how much room your slippage setting leaves before your transaction would revert.
Step three. The bot buys ETH from that same pool immediately before you, pushing the price up. your trade then executes at the degraded price, right at the edge of what your slippage allows.
Step four. The bot sells straight back into the pool imediatly after you, capturing the difference. Three transactions, one block, guaranteed profit. The bot took no market risk at all because it knew both sides before it committed. In any case, The reason this works is that your slippage tolerance tells the bot precisely how much it is allowed to steal..... Set 1% slippage on a thin pool and you have published the maximim bounty. Set it too tight and your transaction reverts and you burn gas for nothing.. That tension is the whole game.
Individual losses are usually smaller than the horror stories suggest... Research analysing private path sandwiching on Ethereum in late 2024 found a median victim loss of roughly $27, a mean of about $137, and a long tail reaching into the low thousands... broader studies covering the erlier public mempool era put average per victim losses far higher, into four figures.... Both can be true. The median trade is small and gets nicked. The occasional large trade on a shallow pool gets filleted.
Who actually gets paid when your transaction is reordered
Strangely, mEV on Ethereum is not one bot in a basement.... It is a supply chain with four distinct layers, and understanding who sits where explains why the probem has been so hard to kill.
Searchers spot the opportunity and build the bundle...... Builders assemble bundles into a full block and bid for the right to have it proposed. Relays sit between builders and validators, holding the block contents in escrow so neither side can cheat the other... validators pick the highest bid and propose it. , This structre is called proposer builder separation, and the software most validators run to participate in it is MEV Boost.The design was meant to democratise access to MEV so that small solo validators earn the same as large staking operations. On that narrow goal it worked.... On decentralisation it did the opposite. As of mid 2026 the block builder makret is severely concentrated, with the top three builder operations responsible for roughly 92% of MEV Boost blocks. Titan alone accounts for over half....... The relay layer is less lopsided, with four relays each holding roughly a fifth to a quarter of payload share. But that is still four entities standing between the worlds transactions and the chain.
This matters beynd MEV itself... A handful of builders deciding what gets included is a censorship surface. And European regulators have already flagged builder and relay concentration as the main structural drawback of the MEV Boost model. the bot taking $27 from your swap is the visible symptom. The concentration is the actual disease.
Why Solana MEV works notihng like Ethereum MEV
Solana has no global public mempool. Transactions are forwarded directly to the current leader, which removes the open marketplace where Ethereum bots hunt. That does not eliminate MEV, it relocates it.
The dominant infrastructure is Jito, whose modified validator client runs on more than 95% of active stake by mid 2026, with Jito tips accounting for over 60% of all priority fee volume on the netwrok. Searchers submit bundles with tips attached, competing in an off chain auction for placement... Bundles execute atomically, which is the important part. Nothing can be inserted between the transactions in a bundle, so a sandwich around a protected trade is structurally impossible rather than merely unlikely.
Jito shut down its public mempool stream in March 2024 specifically becuase it was enabling sandwich attacks, giving up real revenue to do it. Sandwiching on Solana did not vanish afterwards, but it got materially harder and moved to private forwarding arrangements. The newer Block Assembly Marketplace pushes further in the same direction, moving transaction sequencing into trusted execution environments so that ordering happens inside an encrypted enclave that even the scheduler operator cannot inspect.
The trade off is worth naming plainly. Solanas approch reduces sandwiching by making ordering private and verifiable, at the cost of routing most of the networks transaction flow through one dominant piece of infrastructure..... Ethereums approach keeps ordering open and auction based, at the cost of leaving your trade readable by anyone.... Neither has solved it... They have chosen different faliure modes.
Do private RPCs actually protect you from MEV
The standard advice is to route your transactions through a private RPC endpoint so they never touch the public mempool. that advice is broadly correct and worth following, but the providers are not interchangeable, and one difference in particular is rarely mentioned.
Plot twist, an academic benchmark study measured 273 transactions accross the four main protection RPCs on Ethereum, tracking inclusion speed, failure rates and how much of the resulting backrun value was actually returned to the user. The results are uncomfortable reading if you assumed MEV protection meant the value comes back to you.
| Provider | Success rate | Blocks to inclusion | Backrun vlue generated | Returned to user |
|---|---|---|---|---|
| MEV Blocker | ~85% | 1.34 | 0.0035 ETH | 0.0035 ETH |
| Flashbots Protect | ~85% | 1.49 | 0.0088 ETH | 0..0016 ETH |
| Blink | ~84% | 1.56 | 0.0173 ETH | 0 |
| Merkle | ~70% | 1.86 | 0...0108 ETH | 0 |
Read the last two columns again. Two of the four providers generated more backrun vlue per transaction than anyone else and passed none of it to the user who created the opportunity. Your trade is still the raw material... The extraction just happens somewhere you agreed to.
The failure rate difference has a mechanical explanation too. Providers that take a cut out of the priority fee make their transactions less profitable for builders to incldue, so builders deprioritise them. In the studys block space analysis, roughly two thirds of transactions from the two fee skimming providers were excluded on profitability grounds.. slower inclusion is not bad luck, it is the fee model showing up in the data. At the end of the day, Adoption is not niche either. MEV Blocker alone routes something in the range of 10% to 20% of daly Ethereum mainnet transactions depending on the day and the measure used.. A meaningful share of Ethereums order flow now bypasses the public mempool entirely.
Where your transaction is exposed by network
Not every on chain action carries MEV risk, and confusing the ones that do with the ones that do not leads people to worry about the wrong things. our analsis, combining mempool architecture, sequencer design and current protection infrastructure across the major networks, sorts it out.
| Network | Public mempool | Main MEV vector | Sandwich risk on a normal swap | Standard defence |
|---|---|---|---|---|
| Ethereum L1 | Yes | Builder auction, PBS | High wihtout protection | Private RPC, tight slippage |
| Solana | No | Leader forwarding, bundle auctions | Moderate, structurally reduced | Jito bundles, protected RPC |
| Arbitrum | No | Single sequencer, priority auction | Low | Sequencer ordering rules |
| Base | No | Single sequencer | Low | Sequencer ordering rules |
| BNB Chain | Yes | Validator side bundles | High witout protection | Private relay endpoints |
| Plain transfer, any chain | N/A | None | None | Nothing needed |
That last row is the one most people miss. MEV needs a price to move against... Sending USDT from one address to another has no price and no counterparty, so there is notihng to extract. Chains with a single sequencer look safe on this table for now. , But the safety comes from a centralised operator choosing not to extract rather than from the protocol preventing it. Arbitrum has already shipped a timed auction for priority ordering and has signalled a multi party sequencer, which is progress and also an admission that the ordering right is worth money.
What Ethereum is building to fix transacton ordering
Two workstreams matter, and they solve genuinely different problems. People conflate them constantly.
The first is fork choice enforced inclusion lists, specified as EIP 7805 and known as FOCIL, scheduled for inclusion in a coming Ethereum upgrade. Each slot, a committee of around 17 randomly selected validators publishes a list of transactions it has seen and believes should be included. The proposer must includ them or attesters will not vote for the block. This is a censorship fix. It guarantees your transaction gets in..... It does nothing about where in the block it lands, which means it does not stop sandwiching at all.
The second is encrypted mempools, where transactions are submitted encrypted and only decrypted once ordering is already committed. this is the actual sandwich fix, becuase a bot cannot front run a payload it cannot read. It is also much harder, since it introduces new questions about who holds decryption keys, what happens when decryption fails, and how much latency the scheme adds. Proposals exist and are being actively worked on, but this is not a next quarter change.
The honest summary is that Ethereum is closer to guaranteeing your transacton gets included than to guaranteeing it gets included at a fair position. Solana, through encrypted enclave based sequencing, is attacking the ordering problem more directly while accepting more infrastructure centralisation to do it.
How to actually reduce what MEV costs you
Six things, ranked by how much difference they make relative to the effor. Look, Set slippage manually. This is the highest leverage change and it takes five seconds. default slippage on a deep pair is often ten times more generous than the trade needs. On a liquid pair, 0.1% to 0.3% is usually plenty. Your slippage setting is the ceiling on what a sandwich can take, so stop leaving it wide open.Use a protectoin RPC, and check whether it refunds. Routing through a private endpoint keeps you out of the public mempool.. Check the rebate policy before you commit, because the benchmark data shows two of the four major providers keep 100% of the backrun value your trade generates. Mind you, Split large trades. A single large swap on a shallow pool is the most profitable sandwich taeget that exists. Several smaller swaps move the price less per transaction and produce a much thinner margin for anyone trying to wrap them.
Roughly, prefer intent based venues for size... Solvers competing to fill your order off chain and settle in batches remove the ordering advantage entirely, because there is no pending public transaction to front run.
Know when none of this applies. Deposits and withdrawals to and from a platform are transfers, not swaps. There is no price to move, so no MEV. Some platforms lean into this. CryptoCasino.Vegas, for example, settles deposits and withdrawals as direct on chain transfers in the asset you are allready holding.... Which means the blockchain is the only variable and there is nothing in the transaction for a bot to trade against.... The MEV bill, if you paid one, was already paid earlier, at the moment you swapped into that stablecoin.
The part nobody selling MEV protection will tell you
MEV is not a bug that a future upgrade deletes. It is the econmic value of deciding what order things happen in, and any system where a party chooses that order will have it....... You can make ordering private, auction it transparently, redistribute the proceeds, or hide the contents until it is too late to exploit them. You cannot make the ordering right worthless, because it is not.
What you can controle is how much of it you personally hand over.. A default slippage setting, a public mempool and a large market order is a combination that reliably donates money to strangers. Tight slippage, a refunding private endpoint and a split order is not glamorous, and it is most of the available protection.
The bots are not going anywhere. They are just very good at raeding, and the cheapest defence is to stop broadcasting so loudly.
