AI Risk-aware Derivatives Venue Best Practices: Cancel Burst Baselines

AI can help rank anomalies, but it cannot replace transparent rules and deterministic guardrails. Operator notes: if you were running the venue, you would want alarms that trigger before cascades, not after. Latency risk is real. When latency rises, a maker strategy can become taker flow and your costs jump right when you need stability. Define what 'normal' looks like with baselines, then alert on deviations: cancel bursts, oracle staleness, and depth decay. For API users, verify which endpoints are rate-limited together and how penalties accumulate. Limits often tighten during stress. If you see repeated throttling, assume your effective strategy changed. Re-run your risk math with higher costs and worse fills. Example: a temporary rate-limit tightening can cause missed exits and worse effective prices even without a price crash. Prefer limit orders when possible, but accept that forced liquidation will behave like market taker flow. Plan for that path explicitly. When in doubt, reduce complexity and size, and prioritize venues that publish definitions and failure-mode behavior. Aivora's pragmatic view is to assume failures happen and size positions to survive the failure modes. Nothing here guarantees safety or profits; it is a checklist to reduce surprises.