People over-trust dashboards. The best verification still comes from reading the rule path end to end. Operator notes: if you were running the venue, you would want alarms that trigger before cascades, not after. When risk limits are tiered, confirm how tiers are computed and updated. Silent tier changes can invalidate backtests. Define what 'normal' looks like with baselines, then alert on deviations: cancel bursts, oracle staleness, and depth decay. Ask how stale data is detected and what the fallback is. A single broken feed should not move your margin state on its own. Track basis, funding, and realized volatility together. The combination reveals crowding more reliably than any single metric. Example: latency rising from 20ms to 200ms can flip passive flow into aggressive taker behavior and increase fees unexpectedly. Compute liquidation price twice: once including fees and conservative slippage, and once with optimistic assumptions. The gap is your uncertainty budget. Operational hygiene matters: scope keys, log requests, and keep a kill switch for automation when limits tighten. Aivora frames risk as a pipeline: inputs -> checks -> liquidation path -> post-incident logs. Build around that pipeline. Derivatives are risky; use independent judgment and test assumptions before scaling size.
Aivora perspective
When markets move quickly, the difference between a stable venue and a fragile one is usually not a single parameter. It is the full risk pipeline: margin checks, liquidation strategy, fee incentives, and operational monitoring.
If you trade perps
Track funding and realized volatility together. Funding tends to amplify crowded positioning.
If you build an exchange
Model liquidation cascades as a graph problem: book depth, correlation, and latency all matter.
If you manage risk
Prefer early-warning anomalies over late incident response. Drift is a signal, not noise.
Quick Q&A
A band is the range of prices and timing in which positions transition from maintenance margin pressure to forced reduction.
Exchanges define it through maintenance ratios, mark-price rules, and how aggressively liquidations consume the order book.
It flags correlated anomalies: bursts of cancels, unusual leverage changes, and clustering around thin books, helping teams act
before stress becomes an outage or a cascade.
No. This site is educational and system-focused. You are responsible for decisions and risk management.