What "MEV exposure" means in a swap
MEV — maximal extractable value — is value that block builders and searchers can capture by reordering, inserting, or front-running transactions. For a normal swap the practical forms are sandwich attacks and frontrunning, which show up as worse execution than the mid quote suggested. A MEV check estimates how exposed a supported route is to that activity at a point in time.
Routescore expresses exposure as a modeled figure tied to the route and trade size, built from public detector signals. It is a comparison input — high modeled exposure is a reason to look at a cleaner route or a protected submission path, not a prediction that a specific trade will be attacked.
Sandwich and frontrun caveats, stated up front
Modeled exposure is necessarily incomplete. Mempool conditions, builder behavior, and searcher competition change block to block, so the figure is point-in-time and route-scoped. Unsupported routes are not silently scored — they are marked so you never mistake a gap for a low-risk result.
Routescore does not run your transaction through a private relay or change your wallet settings. Where a public protected-RPC option exists for Ethereum mainnet, it is surfaced as external context to review yourself — no route or RPC can promise that a swap avoids MEV.
Why the output is not an execution guarantee
A MEV checker tells you about exposure; it cannot control the block. The honest framing is that Routescore models the risk and compares routes, and you decide how to act in your own venue. There is no avoided-loss or protection claim attached to the number, because realized outcomes are what would justify one — and those are tracked separately through saved records and calibration.
That boundary is deliberate. It keeps the checker useful as pre-decision context without overclaiming what a model can know about an adversarial, fast-moving system.
Turn an exposure check into a record you can review
Save the scenario and the modeled exposure becomes part of a decision record — route, notional, model and feature versions, source freshness, and caveats included. If you later attach an outcome label, that record becomes calibration evidence: it is how the methodology earns or loses credibility over time, in public, instead of by assertion.