Patterns no rule has seen yet.
A model layer sitting beside the rule engine, trained on 3.24M accounts, that surfaces coordinated behaviour, latency abuse and emerging manipulation patterns your written rules were never designed to catch.
Rules only catch what you already know.
Novel abuse
Every rule is written after somebody exploited the gap. The first cohort through a new technique always gets paid.
Coordination is invisible per-account
Ten accounts each trading within the rules can still be one strategy hedged across the firm's book.
Analyst bandwidth
Reviewing thousands of accounts by hand means the team samples, and sampling misses the organised cases.
Detection that improves with the network.
Cluster detection
Identify correlated position-taking across accounts that individually look compliant.
Latency abuse
Flag execution patterns consistent with feed arbitrage rather than discretionary trading.
Cross-account hedging
Detect opposing positions structured to harvest payouts regardless of direction.
Behavioural scoring
A per-account score that moves with behaviour, surfacing accounts worth a human look.
Evidence attached
Every flag arrives with the trades, timings and accounts that produced it, ready for review.
Human in the loop
The layer recommends and evidences. Freezing, failing and paying remain your team's decisions.
How it runs.
What's included
Standard on every firm running on the platform. No add-on tier, no per-seat unlock.
- 01Copy-trade and cluster detection
- 02Latency and arbitrage patterns
- 03Cross-account hedging
- 04Anomalous funding and payout behaviour
- 05Emerging pattern alerts
- 06Model scores exposed over the API
Questions operators ask
Run detection on your history.
Share a historical window and we'll show what the layer would have flagged.