Crypto Prop Risk Architecture (Framework)

A crypto prop risk model converts every position’s characteristics into exposures you can budget. For spot, perps, options, and basis structures, that means delta, gamma, vega, theta, funding sensitivity, correlation, and liquidity impact. The model then aggregates exposures by instrument, sector, and factor, sets hard limits, and defines actions that fire in milliseconds. A desk-grade model is modular: it survives basis shocks, regime changes, and venue outages without a rebuild.

Understanding Crypto Prop Risk Models

A risk model is not a spreadsheet of position sizes. It is an executable system that ingests market data, computes exposures in real time, enforces limits deterministically, and captures every event with an audit trail. The baseline design covers five layers:

• Position analytics: compute Greeks, cash flows, and slippage estimates for each leg of every position.
• Aggregation: build exposure buckets by symbol, sector, and risk factor. Consolidate into a portfolio-level risk view that refreshes continuously.
• Controls: map each bucket to hard limits, soft rules, and deterministic triggers that fire without discretion when thresholds are hit.
• Monitoring: watch limits in real time and capture every breach with immutable audit trails. Nothing passes silently.
• Automation: enforce pre-trade, intratrade, and post-trade behavior without manual intervention when hard thresholds are exceeded.

Architecture and Controls

Assessment

• Greeks and cash flows: compute delta, gamma, vega, theta, and funding sensitivity per contract. Forecast the next 3 funding payments to price carry risk accurately.
• Historical and scenario VaR: use a hybrid approach with at least 2 years of data for majors. Overlay intraday BTC and ETH drop scenarios of 10–20%, funding spikes to 0.1% per 8-hour interval, and 50% liquidity reduction on alt positions.
• Expected Shortfall at 97.5%: VaR alone misses tail losses. ES captures what happens when VaR is breached, critical for flash crash sizing.
• Liquidity and slippage: model from order book depth, spreads, and urgency. BTCUSDT and ETHUSDT spreads run approximately 1–2 bps during liquid hours; alt spreads can be 10–50× wider under stress.
• Correlation and factor risk: cluster assets by L1s, L2s, DeFi, and AI themes. Use regime-switching correlations that widen under stress, static correlation assumptions understate joint drawdown risk.
• Operational risk scoring: score exchange maintenance schedules, API stability, and downtime history. Penalize expected returns during elevated downtime windows.

Controls

• Hard limits: daily and session loss caps, max leverage, position ceilings by symbol and sector. Template: daily realized loss cap at 2% of equity, portfolio hard stop at 4%.
• Pre-trade checks: validate instrument whitelists, leverage caps, margin buffers, and ADV participation before any order reaches the exchange. Reject or auto-resize on any limit breach.
• Server-side stops: place reduce-only stops on entry, priced off mark. If 1h ATR equals 1.5%, do not set a trend stop tighter than 0.75%, tighter stops in that regime produce systematic stop-outs.
• Circuit breakers: at 1.5% intraday drawdown, cut maximum position size by half and block adding to losing positions. On hard cap breach, flatten all positions and pause the session.
• Kill switch: test monthly. Triggers include net loss thresholds, disconnect detection, and volatility z-scores exceeding defined bounds. A switch that has never been tested in production is not a control.
• Hedging rules: if portfolio delta exceeds 0.5 per 10,000 USDT equity, auto-hedge with BTCUSDT or ETHUSDT perps within 60 seconds.

Evaluation and Tooling

Track risk-adjusted returns using rolling 30-day and 90-day Sharpe and Sortino ratios. Add RAROC, net PnL divided by VaR or ES, to measure return per unit of tail risk, not just per unit of volatility.

Core drawdown statistics include max drawdown, average drawdown depth, recovery factor, and time spent in drawdown. These four metrics together indicate whether a system is recovering cleanly or grinding through structural edge decay.

Execution quality tracking covers slippage versus quote, fill ratio, maker/taker mix, and fees as a percentage of gross PnL. A one-click scenario engine should stress test a 15% BTC drop in 30 minutes or a 50% alt liquidity reduction without requiring manual position queries.

Data integrity requires validating pricing and funding data with checksums. Log every pre-trade check, breach, and manual override. Retain at least 12 months of audit data for review and regulatory compliance.

Identifying and Quantifying Key Risks

Market Volatility

Crypto perpetual markets have no daily close. Weekend gaps, thin overnight books, and liquidation cascades that feed on themselves create volatility regimes that equities models do not anticipate.

Regulatory and Compliance

Delistings, KYC tier updates, and listing suspensions force involuntary position changes. Build compliance flags directly into the risk engine so pairs under regulatory review are blocked from new position openings automatically.

On regulatory event days, restrict new position openings entirely or require wider stops and lower leverage as a precondition for entry approval.

Liquidity and Counterparty

Order book depth concentrates on majors. Mid-cap and tail assets thin dramatically during risk-off episodes. Build participation caps relative to 30-day ADV and dynamic minimum book depth thresholds, if the book does not meet the threshold, the pre-trade check rejects the order.

Counterparty risk includes venue stability, liquidation engine behavior under cascade conditions, insurance fund size, and auto-deleveraging mechanics. Maintain documented withdrawal and API contingency procedures for each venue you use.

Funding and Basis Dynamics

Funding is a real cash flow, carry strategies live and die by funding forecasts. Forecast funding across the next 3 intervals. Cap daily funding PnL as a percentage of equity. Define unwind rules that trigger if funding flips or if basis widens beyond a defined threshold.

Operational Risk

API rate limits, WebSocket disconnects, clock drift, and scheduled maintenance create blind spots in position awareness. Run a preflight checklist before each session:

Mitigation and Capital Allocation

Hedging Playbook

Allocation and Throttling

• Volatility scaling: size positions inversely to recent realized vol. If 20-day vol doubles, cut position size in half. This fires automatically, not as a discretionary choice.
• Fractional Kelly: operate at 0.25–0.5 Kelly with drawdown caps. Hard cap per-trade risk at 0.25–0.50% of equity regardless of Kelly output.
• Equal risk contribution: each active strategy contributes the same marginal VaR and ES to the portfolio. Rebalance when contributions drift more than 20% from target.
• Drawdown-aware throttling: at 3% drawdown reduce risk by 30%; at 5% reduce by 60%. Restore to full allocation only after a preset recovery threshold is met, not on a timer.
• Weekend handling: by Friday 22:00 UTC, cut net directional exposure by at least 50% and raise liquidity thresholds for any remaining positions.

Measuring and Monitoring Performance

KPIs That Matter

Real-Time Risk Operations

Live dashboards ingest Bybit WebSocket streams for positions, mark and index prices, funding rates, and order books. Portfolio delta, leverage, VaR, and ES update every second. When any limit is breached, the system triggers automatic hedges or size cuts, senior override is required for manual action during red states.

Post-trade analytics reconcile fills, slippage, and rejects after each session. Compare maker/taker ratios against targets. Flag any strategy where execution quality is deteriorating relative to the prior 30-session baseline, execution decay often precedes strategy decay.

Case Studies from FundedBit

Case 1, Volatility Spike, Momentum System

Context: momentum strategy on BTCUSDT and ETHUSDT perps. 0.5% target stop, 1.2% target profit. Maker-first entries, taker exits. Over 8 days, BTC realized vol rose from 28% to 52% annualized.

Actions taken: the risk engine halved gross exposure and widened the minimum stop by 25% when 1h realized vol exceeded the 90th percentile of the 90-day distribution. Daily realized loss capped at 2%. Circuit breaker at 1.5% intraday drawdown cut new risk by half automatically.

Case 2, Basis Strategy With Protective Overlay

Context: long spot plus short perps across majors to capture positive funding. Target 6–10 bps per day net of fees on BTC and ETH. Real-time monitoring of funding forecasts and scheduled maintenance windows.

Actions taken: funding sensitivity computed per leg with per-symbol caps on carry exposure. Cross-venue hedges disabled automatically during maintenance windows. A 10-delta protective put overlay deployed during a macro risk period.