Hybrid Margin: Cross Efficiency Without Cross Risk

A margin system that recognizes hedges without pooling solvency.

Every perp DEX picks a side. Isolated margin is safe but wasteful. Cross margin is efficient but dangerous. Hybrid Margin gets the benefits of both by separating two things that most designs conflate: recognizing hedges and pooling solvency.


The Capital Efficiency Problem

A trader running a BTC long and ETH short has a natural hedge. In isolated margin, the protocol doesn't care. Each position needs full collateral independently. The BTC leg needs $10k, the ETH leg needs $10k, even though a broad crypto crash would make one profitable while the other loses. Capital sits idle because the system can't see that the two positions offset.

Cross margin solves this by throwing everything into a shared pool. The BTC profit automatically covers the ETH loss. Less collateral, more leverage, better capital efficiency. But the efficiency comes from pooling solvency across markets, and that has a cost that only shows up in stress.

Hybrid Margin separates these two concerns. It recognizes the BTC-ETH hedge at entry through explicit spread credits, reducing the margin requirement without pooling the collateral. Each market keeps its own silo, its own insurance fund, its own liquidation process. The trader gets roughly 80% of cross margin's capital efficiency while keeping every market's failure domain independent.

Figure 1: Solvency Architecture

Isolated Margin

BTC
Silo
ETH
Silo
SOL
Silo
Each market fully independent
Fbtc
Feth
Fsol
Per-market insurance only
ADL (per market, no cross-subsidy)

Cross Margin

Shared Pool
(BTC + ETH + SOL)
Single pool, all markets draw from same resources
Shared Insurance F
Shared insurance fund
ADL (path-dependent: crash order = payout order)

Hybrid Margin

BTC
Silo
ETH
Silo
SOL
Silo
Per-market silos + explicit sweep primitive
Fbtc
Feth
Fsol
Global Overflow F0 (capped, explicit)
ADL (last resort, deterministic)
Isolated walls everything off. Cross pools everything. Hybrid keeps silos but adds a controlled safety net (F0) and explicit cross-market sweeps. Contagion is impossible because the overflow is bounded and published.

The Liquidation Problem

When markets crash in cross margin, the shared pool becomes a race. The first market to liquidate drains the pool. Winners in the second and third markets find nothing left. The outcome depends on which market crashes first, not on whether the trader was right. Change the crash sequence and different people get paid. This is path dependence, and it means payouts are a lottery during the exact moments that matter most.

In isolated margin, there's no contagion, but there's also no safety net, unless there's an insurance fund. If a loser's $50k collateral can't cover the $70k owed to the winner, the winner eats a $20k loss through ADL. No insurance fund from other markets can help because each market is completely walled off.

Hybrid Margin resolves this through a deterministic waterfall. When a market has a shortfall, it draws from four sources in order: the loser's collateral, the market's local insurance fund, a global overflow buffer, and only then ADL as a last resort. Each market resolves independently through the same waterfall regardless of what's happening elsewhere. Crash order doesn't change who gets paid. The global overflow buffer provides the cross-market safety net that isolated margin lacks, without the cross-market contagion that makes cross margin dangerous.

Figure 2: Default Waterfall (what happens when a market has a shortfall)

Cross Margin Waterfall

Shared pool (all trader collateral)Step 1
Shared insurance fundStep 2
ADL (path-dependent)Step 3
The shared pool means Step 1 is a race. First market to liquidate takes collateral that other markets needed. Whoever gets there first wins.

Hybrid Margin Waterfall

Loser's collateral (market-local)Step 1
Local insurance fund FjStep 2
Global overflow F0 (capped, explicit)Step 3
ADL (deterministic, last resort)Step 4
Each step is market-local until F0. The global overflow is the only shared resource, and it's bounded and published. ADL is genuinely last resort.
▶ Try the interactive simulation

The Design Principle

"We recognize portfolio offsets" is a choice in the risk model. "We mutualize solvency across markets" is a choice in the ledger. Most designs treat these as the same thing. Hybrid Margin keeps them separate: risk-aware where it counts, isolated where it must be.

The result is a margin system where hedged portfolios require less capital, liquidation outcomes are deterministic, and no market can silently drain another market's resources. Not by choosing a side between isolated and cross, but by taking the useful part of each and leaving the dangerous part behind.

Figure 3: S/R/M/L Decomposition (the four design levers)

(S) Solvency Domain

Where losses are borne. Per-market silos (fine S) vs shared pool (coarse S). Hybrid keeps fine S: each market has its own ledger, insurance, and liquidation process. No market can silently drain another.

(R) Risk Recognition

Which hedges count. Hybrid grants explicit ICS credits at entry and uses a mobility-masked covariance at maintenance. Only recognizes offsets the protocol can mechanically realize under stress.

(M) Mobility

How equity moves between markets. In cross, it's implicit. In hybrid, it's an explicit sweepUPnL primitive: settle gains, then transfer. Measurable, throttleable, auditable.

(L) Loss Allocation

What happens when close-out fails. Hybrid uses a deterministic waterfall: local Fj, then global F0, then VMGH recovery, then ADL. Payout order is a function of state, not transaction ordering.

Most designs conflate S and R: "we support portfolio offsets" (R) gets implemented by "we pool solvency" (S). Hybrid Margin proves you can have R without compromising S.
Figure 4: Regime Comparison
PropertyIsolated (A1)Cross (A2)Hybrid (A3)
Solvency domainPer-marketShared poolPer-market
Hedge recognitionNoneImplicit (from pool)Explicit ICS + masked covariance
Capital efficiency100% IM~45% IM~63% IM
Equity mobilityNoneImplicit / automaticExplicit sweep primitive
Loss allocationLocal onlyPath-dependentDeterministic waterfall
Cross-market contagionImpossibleBuilt inImpossible (F0 bounded)
Payout determinismYesNo (ordering-sensitive)Yes
Insurance fundPer-market onlyShared (first-come-first-served)Per-market + global overflow
ADL frequencyHigher (no cross-subsidy)VariableLowest (waterfall absorbs most)
Failure modeKnockout risk for spreadsHidden mutualizationGraceful degradation to isolated
Figure 5: Sweep Mechanism (how UPnL moves in Hybrid Margin)

The sweepUPnL primitive is what makes equity mobility explicit and auditable. Rather than letting gains flow implicitly through a shared pool, it settles unrealized profits from one market into transferable collateral and moves them where they're needed.

Market A (BTC)
UPnL: +$5k
Free equity available
sweepUPnL

settle + transfer
Market B (ETH)
UPnL: -$4k
Approaching MM breach
1. Settle BTC gains at conservative settlement mark (with haircut)
2. Transfer realized collateral from BTC silo to ETH silo
3. ETH position cured: equity restored above maintenance margin
The sweep is an explicit, auditable, single-transaction primitive. It converts unrealized gains into transferable collateral under conservative marks. If mobility degrades (congestion, MEV, oracle issues), the system tightens hedge recognition rather than failing silently.

▶ Try the interactive simulation