We tested 11 margin architectures

We tested 11 different margin architecture approaches for our POC clearing engine. Including: SPAN. Basic Cross-margin. Portfolio VaR. Factor-based VaR. SIMM. ML-enhanced.

Four things matter for institutional clearing:

risk accuracy,

regulatory compliance

real-time speed

and ease of integration

Most architectures optimise for one or two. SPAN is universally used but the risk accuracy is poor without significant modification. VaR is accurate but essentially batch-only. ML is fast but no regulator will sign off on a black box.

We landed on ISDA SIMM as the foundation. It works for immediate swaps clearing. It means we follow a standard already integrated across many asset classes and entities.

To make it applicable to the broader market we add factors to bring it inline with VaR.

We added Almgren-Chriss liquidation cost, regime-conditioned risk weights, and nonlinear tail correction for options books.

The POC engine runs at 6,100+ portfolios per second on a single core. Sub-200 microseconds per portfolio. 48 underlyings across 13 SIMM buckets.

#DerivativesClearing #RiskManagement #CryptoInfrastructure

This first appeared on LinkedIn on 24 March 2026. If you want to comment or discuss, that’s the place.