MEV Searcher Strategies

A blockchain block is an ordered list, and whoever decides the order can pay themselves. Buy before a large swap and sell after it; repay a loan the instant it becomes undercollateralised; close a price gap between two pools before anyone else notices. The value available from choosing what goes into a block and in what sequence is called maximal extractable value, and the firms that hunt it are searchers — small, fast teams running bots that scan chain state, assemble a bundle of transactions, and bid for inclusion. The strategies look like classic trading. The economics do not, because a searcher does not trade against a market; they bid in an auction against everyone else who spotted the same thing.

The four things searchers actually do

• Atomic DEX arbitrage. Two pools disagree on a price. Buy in one and sell in the other inside a single transaction that reverts if it is not profitable.
• Liquidations. A lending protocol marks a borrower undercollateralised. Repay part of the debt and take collateral at a protocol-set bonus.
• Sandwiching. Spot a swap with a loose slippage tolerance, buy ahead of it, sell into it. The trader's slippage is the searcher's revenue.
• Non-atomic CEX-DEX arbitrage. The on-chain leg settles in a block; the exchange leg settles immediately. This is the largest category by value and the only one carrying real inventory risk.

The auction sits between you and the money

Since Ethereum's move to proof of stake, most blocks are assembled through proposer-builder separation. A searcher sends a bundle to builders, who compete to assemble the most valuable block; builders submit through relays to the proposer, the validator whose slot it is, who simply signs the highest bid. The searcher pays via priority fee or a direct transfer to the block's coinbase address.

Worked example: an atomic arbitrage

A large buy pushes ETH/USDC on a thin pool to $2,512.50 while a deep 5bp pool still quotes $2,500.00. Depth lets you move about 400 ETH before the gap closes.

• Gross capture: 400 ETH at an average $6.50 of spread = $2,600.
• Pool fees: 5bp on roughly $1,000,000, twice = $1,000.
• Gas: 320,000 units at 15 gwei is 0.0048 ETH = $12.
• Net before bidding: $1,588.

Three other searchers are looking at the identical block state, so this is a sealed-bid auction over the same prize. Bid 90% of it — $1,430 to the builder — and you keep $158, about 6% of the gross. The strategy is not "capture $2,600". The strategy is "capture $158, several thousand times a month, without ever losing an auction you should have won".

Worked example: a liquidation

A borrower has $2,000,000 of WETH collateral against $1,700,000 of USDC debt. Take an 82.5% liquidation threshold and a 5% liquidator bonus, roughly Aave's parameters for WETH.

• Health check: 2,000,000 × 0.825 = $1,650,000 of usable collateral against $1,700,000 of debt. Health factor 0.97 — liquidatable.
• Close factor of 50%: repay $850,000 of USDC, receive collateral worth 850,000 × 1.05 = $892,500. Gross bonus $42,500.
• Fund it with a flash loan at 5bp: $425.
• Sell the WETH back to USDC: 5bp pool fee of $446 plus roughly 10bp of impact, $893.
• Gas: 700,000 units at 30 gwei is 0.021 ETH = $53.
• Net before bidding: $40,683.

Every liquidation bot on the network saw the same oracle update in the same block. Bid 88% and you keep roughly $4,900 for work that looked, on paper, like a $42,500 opportunity.

What erodes the edge

The auction itself. As searcher competition rose, the share retained fell and the share paid to validators rose. Finding an opportunity has never been the scarce skill.

Order flow went private. A large and growing share of retail swap volume now routes through protected endpoints or batch auctions that never touch the public mempool. Sandwiching — historically the easiest strategy to run — is being deliberately engineered out of existence.

Builder concentration. A handful of builders assemble most Ethereum blocks, and exclusive order-flow arrangements mean the best opportunities can be matched internally before any open auction happens. A searcher without those relationships is bidding on the leftovers.

Failure costs. Bundles that lose the auction earn nothing; bundles that land but fail their state checks still burn gas. Both are paid out of the thin slice you keep when you win.

Layer 2 changed the shape. A chain with a single sequencer and first-come-first-served ordering has no public bidding at all — raw latency to the sequencer replaces the auction, which favours a different kind of firm entirely.

Legal exposure is now real. In 2024 US prosecutors charged two brothers over a roughly $25m exploit of Ethereum's block-building pipeline the year before — the first criminal case aimed squarely at this activity. Where clever ordering ends and fraud begins is no longer a purely philosophical question.

In interviews

Separate the two halves out loud: discovery, which is an engineering problem, and inclusion, which is an auction. Name the four strategy families and say which are atomic. Then work a number — the liquidation example is ideal because the bonus, close factor and flash loan fee are all published parameters you can quote and combine on the spot. Land on the economics: the searcher keeps the difference between their costs and the runner-up's, so the durable edges are latency, gas efficiency and private order flow, not cleverness about which pools are mispriced.