hybrid-perps-spec

L6 Risk & Exposure Management

Phase Adjustment Note (v1.2): Net exposure monitoring, hedging engine, and risk reserve are advanced from Phase 3 to Phase 2. Rationale: Once the Phase 2 routing engine goes live, INTERNAL positions generate net exposure immediately, requiring concurrent hedging capability. Routing mode auto-switching and full risk control dashboard remain in Phase 3.

Responsibility Boundary with L3 Internal Order Book

L6 hedging engine and L3 internal limit order queue are completely independent modules:

• L3 Internal Order Book: Manages pending limit order queues for INTERNAL positions, triggers fills at HL prices. Platform is always the counterparty; no user-to-user matching occurs.
• L6 Hedging Engine: Independent of the order book. When net exposure exceeds thresholds, opens reverse positions on HL to reduce risk. Hedge instructions route through L4.

Flow: Order → L3 internal accounting → L6 real-time net exposure aggregation → exceeds threshold → L6 triggers hedge → L4 proxies to HL

Net Exposure (B-Book)

Net Exposure = Platform’s accumulated directional risk from internalized orders.

Formula (INTERNAL positions only)

BTC Net Exposure = Σ(user BTC long INTERNAL notional)
                - Σ(user BTC short INTERNAL notional)

• Positive → Platform is net short (loss if BTC rises)
• Negative → Platform is net long (loss if BTC falls)
• Near zero → Balanced book, low risk

HYPERLIQUID positions are excluded — their P&L flows through HL, no directional risk to the platform.

Net Exposure & Routing Mode Linkage

Real-time net exposure drives routing mode switch recommendations. Thresholds are configurable in the admin panel (see 11-admin.md).

Net Exposure (absolute value)	System Action
≤ Betting mode trigger threshold (default $50K)	Recommend switch to BETTING_MODE / auto-switch if enabled
Between the two thresholds	Stay in NORMAL_MODE; no action
≥ HL mode trigger threshold (default $800K)	Recommend switch to HL_MODE / auto-switch if enabled + hedge reminder

Auto-switch is off by default. Admin must explicitly enable it. When enabled, the system automatically switches mode when thresholds are crossed and sends a notification.

Hedge Reminder in HL Mode

When HL_MODE is activated, the admin panel displays a prominent alert:

⚠️ Net exposure has reached the high-risk threshold. System has switched to Hyperliquid Mode. Please hedge on Hyperliquid immediately!

A Slack notification is also sent to the Risk Manager with the current net exposure value and recommended hedge size.

Hedging Strategy

Single-Asset Net Exposure	Hedge Action
< $100,000	No hedge
$100K – $500K	Hedge 50% on HL
$500K – $1M	Hedge 80% on HL
> $1M	Stop internalization + hedge 80% on HL

Hedge Position Design

• Use low leverage (2x–3x) to avoid hedge position liquidation
• Batch submission (batchModify) to reduce slippage
• Tracked independently; auto-reduce when exposure decreases
• Hedge costs (fee + funding) charged to risk management cost center

Risk Reserve Fund

Funding Sources

Source	Description
Platform initial injection	Pre-launch capital; recommend ≥ $500K
Client loss allocation	20% of each INTERNAL client loss auto-transferred
Drift fallback deduction	Deducted when HL actual loss > Platform calculated loss
Periodic top-up	Replenished from operating profit when below safety threshold

Client Loss Allocation

Per INTERNAL client loss:
  80% → Platform profit
  20% → Risk reserve
(Ratio configurable in admin panel)

Reserve Thresholds

Balance	Action
≥ $500K	Normal operations
$200K – $500K	Reduce internalization (lower routing threshold)
< $200K	Halt all internalization

Risk Monitoring Metrics

Metric	Alert	Critical	Response
Single-asset net exposure	> $500K	> $1M	Stop internalization + hedge
Total internalization exposure	> $2M	> $5M	Reduce internalization
Daily net loss	> $100K	> $500K	Pause internalization
Risk reserve balance	< $500K	< $200K	Reduce/halt internalization
HL account margin ratio	< 300%	< 150%	Top-up / halt HL opens
HL channel latency	> 200ms	> 500ms	Pause internalization
Single-trade drift rate	> 1%	> 5%	Alert / halt HL for this asset
Daily cumulative drift	> $1,000	> $5,000	Manual risk review

Natural Hedging Effect

In the Platform B-book model, when multiple users trade long and short the same asset simultaneously, the platform’s net exposure naturally approaches zero:

User A: buys BTC $50,000 (INTERNAL) → Platform holds BTC short $50K
User B: sells BTC $50,000 (INTERNAL) → Platform holds BTC long $50K
Net Exposure = $50K - $50K = $0 (long/short fully offset)

This is the core profitability condition for the B-book model:

• When balanced, platform has zero directional risk but still earns trading fees and funding fees
• When users collectively lose (statistical edge), platform captures client loss revenue
• The hedging engine only intervenes when positions become imbalanced (net exposure accumulates)

Daily Loss Circuit Breaker

Daily INTERNAL net loss > $100K → Alert
Daily INTERNAL net loss > $500K → Halt all internalization (route everything to HL)

Daily Net Loss = Σ(user gains) - Σ(user losses) (INTERNAL positions, current day)

Phase Delivery Schedule

Deliverable	Phase
Real-time net exposure calculation	Phase 2
Hedging engine (manual + auto-trigger)	Phase 2
Risk reserve basic management	Phase 2
Risk control monitoring metrics + daily loss circuit breaker	Phase 2
Routing mode auto-switching	Phase 3
Full risk control dashboard (Grafana panels)	Phase 3
Hedging strategy optimization (time-sliced execution, slippage optimization)	Phase 4