An invisible tax on every swap
Swap tokens on a DEX and your transaction sits in a public mempool before it gets included in a block. During that window, actors called searchers can read your pending trade and everything about it, including the maximum slippage you told the wallet you would accept. If your trade is big enough relative to pool liquidity, a searcher can sandwich you: buy just ahead of you to push the price up, let your order fill at the worse price, then sell right after into the move you created.
The math is not complicated. Say you submit a swap to buy ETH with 1% slippage tolerance. A searcher sees it, buys ETH just before you and pushes the price up by 0.8%, your trade fills 0.8% worse than the quote, and the searcher sells into the higher price your own order made. They keep the difference minus gas. You walk away with fewer tokens than you would have gotten in a world without MEV.
Flashbots and EigenPhi data put cumulative MEV extraction on Ethereum alone north of $600 million, and that undercounts it. It only captures what is visible on-chain, so it misses MEV on L2s, on alt-L1s, and cross-domain MEV entirely.
How a sandwich actually gets built
A sandwich needs three things to line up: spotting a profitable target in the mempool, building the front-run and back-run transactions, and getting the whole bundle included in the right order inside one block. Searchers bid for that inclusion through mechanisms like Flashbots MEV-Boost, submitting transaction bundles to block builders who assemble blocks for validators.
Not every swap gets sandwiched. Searchers only bother when expected profit beats the cost of pulling it off, which is gas on three transactions plus the tip to the builder. Small swaps on liquid pairs usually are not worth it. Large swaps on thin pairs, or swaps with wide slippage, are the good targets.
The threshold moves with gas prices and pool depth, but as a rough feel, swaps above $10,000 to $20,000 on moderately liquid pools get hit often. On genuinely thin pools, much smaller swaps are worth sandwiching.
Beyond sandwiches
Sandwiches are the most visible flavor of MEV, not the only one.
- Liquidation MEV. Searchers race to liquidate undercollateralized positions on lending protocols. Whoever lands their liquidation first earns the bonus, often 5 to 10% of the collateral. This one is actually good for protocols, since it keeps liquidations prompt and protects solvency.
- Arbitrage MEV. A searcher notices ETH is $3,500 on Uniswap and $3,505 on SushiSwap, buys on one and sells on the other, and pockets the $5. Also broadly healthy, because it keeps prices consistent across venues.
- JIT (just-in-time) liquidity. Subtler. A searcher sees a large swap coming, drops concentrated liquidity right at the price range the swap will hit, collects the trading fees, and pulls the liquidity back out immediately after. That fee revenue would have gone to passive LPs, and it gets redirected to the JIT provider instead.
What retail can actually do
The single best defense against sandwiches is private submission. Flashbots Protect, MEV Blocker, and various wallet-integrated options route your transaction straight to block builders through private channels and skip the public mempool. If searchers never see the trade, they cannot sandwich it.
Your slippage setting matters more than people think. Set 5% on a swap and you have told the whole network you will accept a price up to 5% worse than quoted, and a searcher will take as much of that 5% as is profitable. Drop it to the minimum that still fills, often 0.5 to 1% on liquid pairs, and there is far less to extract.
Timing and sizing help too. Breaking one large swap into several smaller ones cuts the per-transaction opportunity, though you pay more gas. Trading when the mempool is quieter, which usually means cheaper gas, thins out the searcher competition and the premium you eat.
Aggregators like 1inch or CoW Swap are worth a look because they route across pools and lean on private order flow. CoW Swap in particular runs a batch auction, matching trades off-chain and settling in batches instead of one after another in the mempool, so it is structurally hard to sandwich in the first place.
Where this is heading
MEV is baked into any public blockchain with a transparent mempool and permissionless block production. It is not going away. The real question is whether the extraction is mostly the harmful kind (sandwiches on retail) or the useful kind (arbitrage that tightens prices, liquidations that keep protocols solvent).
The infrastructure is slowly getting better at splitting those two apart. Flashbots MEV-Share hands some of the MEV back to the users whose transactions created it, and order flow auctions let users capture it themselves by auctioning the right to execute their trades. None of this is finished, but for now the practical move is simple. Route your swaps privately, keep your slippage tight, and you take yourself off the menu.