How integrity detection works
This page is deliberately less specific than the others in this section. Its subject is the protocol's fraud and abuse detection, and publishing exact thresholds or rule internals would hand a roadmap to the people it exists to catch. What follows describes what the system does and why it's structured the way it is, not how to evade it.
It watches patterns, not just single accounts
An account trading alone in isolation is easy to evaluate honestly on its own terms. The harder problem is coordinated abuse: one person or group operating multiple accounts to game an evaluation's rules rather than actually trading well, for example by taking opposite positions across linked accounts so one side is guaranteed to pass regardless of the market. Catching this requires looking across accounts and across firms, not just within one, which is why this system operates at the protocol level rather than being something each firm runs independently.
It's a real signal inside the risk engine, not a bolt-on
Elevated abuse activity feeds directly into a firm's own health score (see How the risk engine thinks) as one of its inputs. This is deliberate: a firm with a real, unaddressed abuse problem should show up as a firm under more scrutiny, not treated identically to one that doesn't have one.
The consequence is a rate limit, not a permanent record
A wallet that trips a genuine concern drops into a stricter reputation standing, see Reputation across firms, which slows how fast that wallet can extract funds and shrinks how many accounts it can hold at once, across every firm, not just the one where it was . This is recoverable by design: it isn't a conviction or a ban, and a wallet's standing recomputes fresh once the underlying concern is no longer supported by its actual behavior.
Why fragmenting into new wallets doesn't help
Because detection and its consequences both operate on identity signals that persist across a single wallet's history rather than resetting per account, spreading activity across many fresh wallets to dodge scrutiny doesn't actually escape it, it just means none of those wallets individually build up enough of a track record to earn the looser terms a real, continuous history unlocks. The system is built around making one honest identity more valuable over time than many disposable ones.
What this is, and isn't, a substitute for
This system catches coordinated, pattern-level abuse. It isn't what makes individual fills fair, that's a property of the execution engine itself (see How fills are actually fair), and it isn't what makes a settlement provable, that's the fault-proof system (see How settlement actually works). It's the layer that exists specifically for the abuse patterns those other systems aren't designed to catch on their own.