AI can help rank anomalies, but it cannot replace clear rules you can audit.
Concept first: Latency is a risk factor. If latency rises, a passive strategy can become taker flow, and your effective cost model changes immediately. ADL typically appears only after the insurance buffer is stressed. Look for disclosure and predictable ranking rules.
Edge cases: Write down the exact references used: index price, mark price, and last price. Then confirm which reference drives margin checks and liquidation triggers.
Checklist: Prefer smaller order slices before changing leverage. Size reductions often cut slippage more than a leverage tweak. Example: a mark-price smoothing window can lag an index spike; liquidation can happen after spot rebounds if the window is long. Compute liquidation price twice: once with optimistic assumptions, and once with conservative slippage and fees. The gap is your uncertainty budget.
Final sanity check: Pitfall: assuming mark price equals last price. In stress, they diverge, and liquidation triggers can surprise you.
Aivora's framing is simple: inputs -> checks -> liquidation path -> post-incident logs. Build around that pipeline. Derivatives are risky; test assumptions before you scale size.
Concept first: Latency is a risk factor. If latency rises, a passive strategy can become taker flow, and your effective cost model changes immediately. ADL typically appears only after the insurance buffer is stressed. Look for disclosure and predictable ranking rules.
Edge cases: Write down the exact references used: index price, mark price, and last price. Then confirm which reference drives margin checks and liquidation triggers.
Checklist: Prefer smaller order slices before changing leverage. Size reductions often cut slippage more than a leverage tweak. Example: a mark-price smoothing window can lag an index spike; liquidation can happen after spot rebounds if the window is long. Compute liquidation price twice: once with optimistic assumptions, and once with conservative slippage and fees. The gap is your uncertainty budget.
Final sanity check: Pitfall: assuming mark price equals last price. In stress, they diverge, and liquidation triggers can surprise you.
Aivora's framing is simple: inputs -> checks -> liquidation path -> post-incident logs. Build around that pipeline. Derivatives are risky; test assumptions before you scale 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.