Mapping the MEV landscape

When I started working at Flashbots, I wanted to understand MEV beyond the familiar examples of arbitrage and sandwiching. I kept finding strategies that looked different on the surface but were driven by the same underlying advantage: control over when a transaction executes and what executes around it.

This is the map I built while going deeper into that landscape. It is not a list of guaranteed ways to make money. It is a way to classify where extractable value comes from, who can capture it and what risks are hidden inside each strategy.

What counts as MEV?

Maximal extractable value is value available through the inclusion, exclusion or ordering of transactions during block production. The definition becomes broader when several blocks or execution domains are involved.

Not every profitable onchain action is MEV. A trade is more clearly MEV when its profitability depends on privileged ordering or a specific position around another state transition.

I found four questions useful when examining an opportunity:

Source: What state change creates the value?
Position: Must the transaction execute before or after something else?
Atomicity: Can every leg succeed or revert together?
Externality: Does the strategy improve market state, compete for existing value or directly worsen another user's execution?

1. Atomic DEX arbitrage

The simplest example is a price difference between two onchain markets.

buy on the cheaper pool        ↓sell on the more expensive pool        ↓repay any temporary capital        ↓keep the remainder

If every leg executes inside one transaction, the searcher can revert when the final balance is not profitable. That makes execution atomic, though discovery, inclusion and competition still carry costs.

Arbitrage helps bring onchain prices back together. The opportunity is highly competitive because several searchers can observe and simulate the same state.

2. CEX–DEX and cross-venue arbitrage

A DEX price may diverge from a centralized exchange or another offchain reference market. A searcher can trade onchain while hedging or carrying inventory on the external venue.

This is not the same as atomic DEX-to-DEX arbitrage. The centralized exchange leg cannot normally be included in the Ethereum transaction, so the strategy has latency, inventory, fill and counterparty risk.

The advantage may come from faster market data, faster simulation, better inventory placement or a stronger relationship with order flow rather than from a single smart-contract trick.

3. Backrunning state changes

A backrun executes immediately after a transaction that leaves the blockchain in a predictable state.

The trigger could be:

A large swap that moves an AMM price.
An oracle update that makes a position liquidatable.
A protocol rebalance or asset conversion.
A stablecoin mint, redemption or depeg event.
Another searcher's liquidation that creates a secondary arbitrage.

Backrunning is a broad pattern rather than one strategy. The first transaction creates a state transition and the backrun captures value available in the resulting state.

Bundle composition makes this relationship explicit. An ordered bundle can place a searcher's transaction after another transaction while enforcing execution conditions. Systems such as MEV-Share can selectively reveal hints that let searchers compete for backruns and return part of the extracted value to the originating order flow.

4. Liquidations

Lending protocols need unhealthy positions to be closed before their debt becomes uncollectable. A liquidator repays some debt and receives collateral plus an incentive defined by the protocol.

The MEV opportunity comes from being the transaction that executes the valid liquidation. Searchers monitor positions, collateral prices, interest and oracle updates, then simulate whether the liquidation reward covers execution costs and market risk.

Liquidations serve a protocol function by maintaining solvency. The surrounding race can still create duplicated computation, failed transactions and concentrated advantages for participants with lower latency or better order flow.

5. Frontrunning and generalized copying

A frontrun takes a position before a known transaction because the earlier position is more valuable. The searcher may submit a transaction with stronger inclusion incentives or send an ordered bundle to a builder.

A generalized frontrunner goes further: it detects another profitable transaction, simulates a modified version and attempts to capture the same opportunity first. Private submission channels and bundles reduce the information exposed to this kind of copying.

Frontrunning is a position in the ordering. It becomes a specific strategy only when paired with a source of value such as an arbitrage, mint or auction.

6. Sandwiching

A sandwich surrounds a user's trade:

searcher buys first        ↓user's swap moves the price        ↓searcher sells afterward

The first leg changes the price against the user's trade. The user's own price impact then helps make the final leg profitable. Slippage tolerance places a boundary on how far the transaction can be moved before it reverts.

Unlike an arbitrage contained in one transaction, the user's transaction is a separate signed action. A bundle can enforce ordering, but the strategy still depends on the user transaction executing with the expected state and constraints.

Sandwiching is usually classified as harmful MEV because value is extracted by worsening the user's execution.

7. Just-in-time liquidity

Concentrated-liquidity AMMs make another ordering possible:

Add narrowly targeted liquidity immediately before a large swap.
Earn a large share of the swap fee.
Remove the position immediately afterward.

This is called just-in-time, or JIT, liquidity. It resembles a sandwich around the swap, but the surrounding actions change liquidity rather than trading directionally against the user.

Its effect is more nuanced than a conventional sandwich. JIT liquidity can improve the trader's execution by adding depth while shifting fee revenue away from passive liquidity providers. Research on Uniswap v3 found that it was relatively rare in the measured period and could offer price improvement to the targeted trader.

8. NFT, mint and auction MEV

The same ordering advantage appears outside fungible-token swaps:

Buying a mistakenly underpriced NFT before another buyer.
Obtaining an early position in a limited mint.
Atomically buying several floor listings.
Bidding or settling around an onchain auction boundary.
Capturing a rare trait before metadata or market prices adjust.

These opportunities are less uniform than DEX arbitrage. Token uniqueness, marketplace rules, royalties and uncertain resale value make many of them speculative rather than atomic.

9. Cross-domain MEV

An execution opportunity may span Ethereum, a rollup, another chain, a bridge or an offchain venue. The value then depends on state and ordering across more than one domain.

A simple price difference is only the starting point. Cross-domain execution must account for:

Different finality assumptions.
Bridge or messaging delay.
Capital pre-positioned on each domain.
Reorg and sequencer risk.
The inability to make every leg atomic.
Competition from actors observing both domains.

The important distinction is that cross-domain MEV is not merely “bridge an asset and sell it elsewhere.” Slow bridging may remove the opportunity before capital arrives. Practical strategies often rely on inventory already placed on both sides and rebalance later.

10. Sequencer and latency MEV

On a rollup or another ordered execution domain, a sequencer may control arrival rules, ordering and inclusion timing. Even without reading private transaction content, priority lanes and timing rules can create advantages for participants willing to pay for earlier execution.

This category includes:

Priority or express-lane auctions.
Ordering around sequencer batches.
Delayed inclusion or censorship for economic advantage.
Latency races around public order flow.

The exact opportunity depends on the domain's sequencing policy. A single-sequencer rollup, shared sequencer and Ethereum L1 builder market do not expose the same powers or guarantees.

11. Intent and solver MEV

An intent describes an outcome—such as exchanging an asset above a minimum amount—without prescribing every execution step. Solvers compete to find routes, internal matches and bundles that satisfy it.

Solver competition does not make MEV disappear. It changes where discovery and ordering happen and how surplus is divided among the user, solver, order-flow provider, builder and proposer.

Possible sources of solver value include:

Matching compatible user orders directly.
Routing across several liquidity sources.
Backrunning the resulting settlement.
Combining several intents into one batch.
Using inventory to avoid an unnecessary onchain swap.

Batch auctions and order-flow auctions can return more value to users, but their guarantees depend on auction design, information revealed to solvers and how winning execution is selected.

12. Builder-level optimization

A block builder sees transactions and bundles that may conflict with one another. The builder's problem is not simply “sort by stated profit.” It must find a valid block whose total payment and execution value are highest under constraints such as gas, transaction nonces, state conflicts and bundle ordering.

public transactionsprivate transactionssearcher bundlesconflicting state transitions          ↓simulate and select a compatible set          ↓produce the highest-value valid block

This is a layer of MEV allocation rather than a single searcher strategy. MEV-Boost provides access to a competitive builder market; it is infrastructure for block production, not itself an MEV opportunity.

13. Multi-block MEV and time-bandit attacks

Some value depends on control over several blocks. A block producer might delay an action, condition one block on another or try to replace recent history to capture unusually large MEV already taken by someone else.

The reorg form is commonly called a time-bandit attack. It matters because sufficiently large MEV can create an incentive to challenge consensus stability.

The old shortcut that a reorg simply requires “51% hash power” is incomplete and does not describe proof-of-stake Ethereum. Feasibility depends on proposer timing, stake, fork choice, finality, coordination, detection and the cost of failure. This is a consensus-security risk rather than a normal searcher trade.

14. Protocol-specific long-tail opportunities

The largest research surface is the long tail. Any predictable state transition can create an ordering dependency:

Rebase and index updates.
Fee-on-transfer or unusual token accounting.
Peg maintenance and redemption queues.
Staking deposits, exits and reward updates.
Onchain auctions and batch settlement boundaries.
Vault rebalances and strategy migrations.
Rounding behaviour or abandoned balances that a protocol intentionally makes claimable.

These are not automatically separate MEV categories. They become MEV when inclusion or ordering is necessary to capture the value. A protocol bug that lets anyone drain funds is primarily an exploit even if a bundle, flash loan or frontrun is used to execute it.

What should not be confused with an MEV strategy?

Several items in early MEV lists are better treated as tools or adjacent risks:

Flash loans provide temporary capital. Arbitrage, liquidation or manipulation is the strategy using that capital.
MEV-Boost connects validators to a builder market. It changes the block-production pipeline rather than creating one trade.
Oracle manipulation is often a protocol-security exploit. Backrunning a legitimate oracle update is a clearer MEV pattern.
Flash-loan governance attacks exploit weak voting or timelock design. Ordering may assist the attack, but it is not enough to classify the whole exploit as MEV.
Dust collection is only extractable value when protocol rules leave claimable balances and ordering determines who receives them.

This distinction matters because otherwise “MEV” becomes a label for every profitable or malicious onchain action.

A practical research model

For any proposed strategy, the gross opportunity is not the profit:

expected profit  = execution proceeds  - capital cost  - gas used  - builder or priority payment  - hedge and bridge costs  - expected loss from failure or reorg

The research loop is usually:

Identify the state transition that creates value.
Reconstruct the state needed to simulate it.
Calculate the valid ordering and atomicity requirements.
Include every execution and failure cost.
Decide what information can safely be revealed.
Submit through the appropriate public or private path.
Measure landed transactions rather than simulated profit alone.

The map I want to remember

Family	Typical examples	Main dependency
Price convergence	DEX–DEX and CEX–DEX arbitrage	Price difference
State-transition backruns	Swap, oracle update and rebalance	Position immediately after an event
Protocol maintenance	Liquidations and peg operations	Permissionless incentive
User-order extraction	Frontruns and sandwiches	Position around user flow
Liquidity positioning	JIT liquidity	Position around a swap
Domain-specific markets	NFT mints and auctions	Scarcity and ordering
Cross-domain execution	Rollup, chain and venue arbitrage	Joint state across domains
Order-flow markets	Intents, solvers and auctions	Information and allocation rules
Block production	Bundle merging and builder optimization	Valid block composition
Consensus-level extraction	Multi-block and time-bandit MEV	Control across blocks

The most useful conclusion from this research was that MEV strategies are not a fixed checklist. They are instances of the same deeper problem: a state transition creates value and some participant has an advantage in observing, ordering or executing around it.