A lot of losses come from tiny assumptions: which price triggers liquidation, when funding hits, and how fees are applied.
What it is: Write down the exact references used: index price, mark price, and last price. Then confirm which reference drives margin checks and liquidation triggers.
What to check: An AI risk layer should be explainable: it can rank anomalies, but deterministic guardrails must remain stable and auditable.
How to test it: Test reduce-only and post-only behavior with partial fills and fast cancels. Edge cases often appear during rapid moves. Example: a temporary rate-limit tightening can cause missed exits and worse fills even without a dramatic price crash. Run a small-size rehearsal when liquidity is thin. Observe how stop orders trigger and how mark/last prices diverge around spikes.
Common pitfalls: Pitfall: trusting a single data source. One stale oracle feed can distort index and mark calculations if fallbacks are weak.
Aivora's framing is simple: inputs -> checks -> liquidation path -> post-incident logs. Build around that pipeline. Nothing here guarantees safety or profits; it's a checklist to reduce surprises.
What it is: Write down the exact references used: index price, mark price, and last price. Then confirm which reference drives margin checks and liquidation triggers.
What to check: An AI risk layer should be explainable: it can rank anomalies, but deterministic guardrails must remain stable and auditable.
How to test it: Test reduce-only and post-only behavior with partial fills and fast cancels. Edge cases often appear during rapid moves. Example: a temporary rate-limit tightening can cause missed exits and worse fills even without a dramatic price crash. Run a small-size rehearsal when liquidity is thin. Observe how stop orders trigger and how mark/last prices diverge around spikes.
Common pitfalls: Pitfall: trusting a single data source. One stale oracle feed can distort index and mark calculations if fallbacks are weak.
Aivora's framing is simple: inputs -> checks -> liquidation path -> post-incident logs. Build around that pipeline. Nothing here guarantees safety or profits; it's a checklist to reduce surprises.
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.