MEV research: trace the user's execution outcome
MEV research should connect transaction ordering to an observed user outcome and an explicitly defined comparison. A suspicious-looking sequence is a lead, not a complete measure of harm. Record the user’s actual asset flows, the relevant surrounding transactions and the assumptions behind any counterfactual execution price.
Ethereum’s MEV overview explains that some ordering strategies can worsen execution for users. That general mechanism does not identify the cause of every unfavorable fill. This worksheet studies evidence and attribution; it does not provide instructions for extracting MEV or manipulating another user’s transaction.
Begin with realized receipts
Identify the transaction, network, block and asset pair. Measure compatible input units, output units, explicit fees and any token-level deductions. Use the actual receipt rather than an interface’s earlier indicative quote. Quotes can change for reasons unrelated to the ordering pattern under investigation.
Next locate relevant trades before and after the user’s transaction. Preserve pool identities and state transitions. Adjacency alone does not prove common control or an intentional strategy. Labels from an analytics service should be retained as attributed classifications unless independently supported.
Separate total slippage from the component assigned to the suspected ordering. Price movement, trade size, route selection, liquidity changes and token behavior can affect receipts. A poor fill does not automatically reveal which mechanism caused it. A causal statement requires more than comparing a quote from several minutes earlier with the final output.
Hypothetical execution comparison
Assume a fictional user pays 1,000 units of A and receives 480 units of B. An earlier interface quote showed 500 B. The visible shortfall is 20 B, but that number alone is not a measured MEV loss. The quote may precede unrelated pool changes.
Suppose a carefully specified reconstruction excluding two identified surrounding transactions estimates 492 B at the user’s position in the block. Under that counterfactual, the modeled difference is 12 B. The remaining eight between 500 and 492 belongs to other changes or assumptions, not automatically the suspected strategy.
The reconstruction must state how it handles other transactions, gas, route behavior and failed executions. Removing a transaction can alter subsequent states, so the result is model-dependent. All numbers are hypothetical; no real experiment or attack attribution is claimed.
Outcome attribution worksheet
| Layer | Observation or assumption |
|---|---|
| Actual receipt | Verified input, output and fees |
| Earlier quote | Timestamp, route and indicative conditions |
| Surrounding sequence | Transaction identifiers and affected pools |
| Classification | Evidence for the suspected pattern and controller |
| Counterfactual | Removed events, state model and replay assumptions |
| Estimated difference | Output delta with uncertainty and exclusions |
Write an observation sentence separately from an inference sentence. For example, transactions occurred on both sides of the user trade is an observation; they caused the entire quote shortfall is a stronger inference. A table should not silently transform the first into the second.
Bound the conclusion
If no credible counterfactual is available, report the observed sequence and receipt without assigning a precise loss. If several plausible reconstruction rules produce different deltas, show the range and explain the choices. A precise-looking number from one replay configuration is not automatically a uniquely identified economic effect.
For spot protocol research, ask whether execution quality evidence is broad enough to support a protocol-level claim. One example can illustrate a failure mode but cannot establish population frequency or average user cost. Sampling, measurement coverage and period selection must be addressed before generalizing. Avoid turning a single classified incident into a price prediction or a claim about all users.
FAQ
Is every unfavorable fill evidence of MEV harm?
No. Establish the relevant ordering mechanism and distinguish it from price movement, liquidity and route effects.
Can I call the difference from an old quote the MEV loss?
Not without a justified comparison. The quote and execution may differ in timing and state for unrelated reasons.