Open-Source Risk Engine (ORE) and Risk Analytics Lab: Aligning XVA Analytics with industry best practices

In the wake of post-2008 regulatory reforms and escalating computational demands, banks have sought more transparent and cost-effective solutions for valuation adjustments (XVAs). The Open-Source Risk Engine (ORE) emerged in 2016 as a free, peer-reviewed risk analytics platform built on QuantLib's foundations. It provides a transparent, extensible framework for pricing and risk analysis across a broad array of analytics spanning market risk and counterparty credit risk, offering an open alternative to proprietary risk engines. ORE is actively maintained and forms the core of Post Trade Solutions’ (formerly Acadia's) Risk Suite for Uncleared Margin Rules, a critical industry resource for the non-cleared derivatives market. The platform has been adopted by a broad range of institutions, processing hundreds of thousands of trades per day in production services. Complementing ORE's on-premises deployment, the new Risk Analytics Lab (RAL) delivers the same capabilities via a cloud-based SaaS platform. This dual operating model – local ORE and hosted RAL – ensures that whether firms are traditional in-house or cloud-first, they have a clear path to adopt modern XVA analytics without compromise.

ORE extends QuantLib with comprehensive support for XVA and risk simulations. Unlike QuantLib's library-only approach, ORE provides out-of-the-box portfolio analytics such as collateralized pricing, CVA, DVA, FVA, MVA, KVA, exposure simulation, Value-at-Risk and stress testing. It also expands product coverage and introduces "scripted" trades (flexible payoff definitions for complex structures). These enhancements turn the QuantLib base into a full risk engine usable directly by front-office, risk, and validation teams. Equally important is ORE's modelling approach, which aligns closely with prevailing industry practices. As documented in Risk.net's XVA benchmarking series ("Most banks stick to tried-and-trusted XVA models", Fabio Santos, November 18, 2025), ORE's core stochastic models mirror these widely adopted approaches, providing familiarity, scalability, and validation continuity while maintaining a structured path toward richer dynamics. 

Interest Rates: ORE’s exposure simulation uses the Hull-White family of models (also known as Linear Gauss-Markov). This falls under the Vasicek/HW framework that remains common in production XVA stacks, largely for scalability and governance reasons.  A recent 30-bank survey found 80% of desks use HJM variants, with 59% specifically relying on one-factor short-rate models like Hull-White - a common baseline driven by tractability, but one with well-known limitations that are motivating a shift toward multi‑factor and richer dynamics. ORE supports this baseline for adoption and validation continuity and provides a controlled path to uplift via multi‑factor Hull‑White and related enhancements.

Equities and FX: For equity and foreign exchange exposures, ORE employs lognormal diffusion (Black-Scholes) models. This choice reflects industry standards: 65% of XVA desks use Black-Scholes for equity underlying models, and 73% do so for FX. ORE’s analytics leverage these tractable models for simulation and options pricing, ensuring consistency with the primary approach used across the market. A sizeable minority of banks – including some large dealers – augment Black-Scholes with more advanced local or stochastic volatility models for certain portfolios. While ORE’s current release focuses on the mainstream lognormal models, its roadmap (see below) includes integrating local volatility and Heston stochastic volatility, extending coverage to those more advanced practices as well.

Credit: Credit in ORE is supported by intensity-based models for counterparty default risk. This includes simulating default times from hazard rates (calibrated to credit spreads) using processes like CIR++ (a shifted Cox-Ingersoll-Ross model) or analogous short-rate models applied to credit curves. Industry benchmarking shows this approach is prevalent: more than two-thirds of XVA desks use intensity-based credit models calibrated to term structures of credit spreads. ORE also permits simpler deterministic credit exposure methods (e.g. using static credit curves for default probability), though like in industry, the intensity-based stochastic simulation is the workhorse.

Commodities: ORE provides commodity risk factor models (such as one-factor Gabillon or Schwartz models) for xVA on commodity trades. Risk.net’s study indicates commodity exposures are often treated with either Black-Scholes (48% of firms) or firm-specific bespoke models – ORE’s current commodity module falls into this “Black-Scholes” category. Plans are in place to generalize this to multi-factor commodity processes, as discussed later.

The table below summarizes how ORE’s methodologies map to the XVA model categories identified by Risk.net, highlighting that ORE supports the core models prevalent in practice:

Table 1: ORE models mapped to Risk.net XVA model categories.

Notably, ORE’s emphasis on well-established analytical models echoes the broader industry trend. According to Risk.net’s XVA benchmarking, banks “overwhelmingly favor tractable analytical models” for XVAs and strive to use the same models as their trading desks for consistency. Only a handful of firms have experimented with esoteric approaches like copula-based wrong-way risk simulation. ORE’s design philosophy is similar: cover the fundamental, trusted models first – ensuring transparency and ease of validation – while allowing extensibility for advanced methods as they mature. This gives institutions confidence that adopting ORE won’t put them out of step with mainstream risk management practices.

A key strength of ORE is the flexibility in how it can be deployed. Traditional users can integrate ORE on-premises as part of their internal risk infrastructure, embedding the engine into existing workflows and systems. At the same time, the Risk Analytics Lab (RAL) offers a cloud-native deployment for ORE’s analytics, delivered as a fully managed service. RAL combines ORE’s open-source engine with on-demand scalable computing, managed configuration, data integration, and a user-friendly interface. Delivered through the SaaS platform, RAL requires no installation, no lengthy integration, and no upgrades – users can launch the environment instantly via the web and start analyzing risk from day one.

In practice, RAL provides a Jupyter/Python-based analytics workspace on the cloud, pre-loaded with ORE’s libraries and 15+ years of cleansed market data for calibration and backtesting. Users can run complex pricing models for XVA, Historical VaR, stress testing, and model validation in this environment. The platform’s design supports scalable computation – jobs are queued and executed on cloud infrastructure, with the ability to autoscale for heavy workloads and even continue long-running tasks after the user logs out. Through intuitive notebook-based scripting (via Python APIs to ORE), analysts can customize workflows and automate runs (scheduled or event-driven) without worrying about managing servers. In short, RAL delivers ORE’s full analytical power “as a service,” backed by an enterprise-grade cloud setup and integrated data sources.

This dual delivery model directly addresses the diverse preferences seen across the industry. Many banks are increasingly open to cloud deployment for XVA: nearly two-thirds of XVA desks now run at least part of their workload in the cloud, and 46% of firms actively tap public cloud resources for compute scaling. In fact, five of the surveyed banks have moved their entire XVA stack to public cloud, and several others use private clouds or hybrid setups. At the same time, 39% of firms still run all XVA calculations on on-premises servers only, due to concerns around data control, latency, or cost predictability. ORE and RAL together offer a solution for both groups: firms can keep using ORE in-house if they require or migrate to RAL’s cloud-native service (now or in the future) without switching analytics frameworks.

Crucially, ORE and RAL are interoperable and consistent. The analytics and models are the same in both environments, so a valuation or risk metric computed in local ORE will match that in RAL. This means a bank could, for example, develop and validate their XVA models on a local ORE installation, and later decide to offload large-scale production runs to RAL’s cloud with minimal friction. Such flexibility aligns with how institutions are handling XVA workloads in practice: many keep day-to-day pricing runs on internal grids (to stay “close to the data and traders”) but burst to cloud for one-off heavy simulations or stress tests. RAL facilitates this hybrid approach by providing on-demand computing power exactly when needed, while ORE’s availability for on-prem use ensures low-latency and full control for business-as-usual calculations. In either case, clients are using the same models, inputs, and codebase, easing validation and governance. As of late 2025, roughly 70% of surveyed banks are either already running XVA in the cloud or actively migrating toward cloud-native architectures – a trend that RAL supports out-of-the-box. For the cohort who remain on internal infrastructure, ORE offers a robust solution that can slot into their existing tech stack. This dual offering instils confidence that an investment in ORE today will not limit a firm’s deployment options tomorrow.

ORE’s development roadmap is driven by community input and evolving market needs, with a focus on expanding model coverage and computational performance in a transparent, non-proprietary manner. The following key enhancements are in progress or planned, with clear distinction between live functionality and forthcoming capabilities:

Interest Rate Models: Multi-factor Hull-White. ORE's v15 release introduced full support for Hull-White models with multiple factors (n-factor Hull-White), including calibration to historical rate curve moves. This allows users to directly set or calibrate mean-reversion and volatility parameters from historical principal component analysis or market option data, providing richer dynamics under the familiar Vasicek/Hull-White framework for interest rate simulations. ORE is committed to adding a Quadratic Gaussian rates model by end‑2026. This provides a more expressive evolution of rates than purely linear‑Gaussian dynamics, improves the ability to represent richer distributional features in exposure simulation, and serves as a structured step toward Cheyette‑type frameworks over time.

1. FX and Equity Models: Local Volatility and Stochastic Volatility. Work is underway to incorporate local volatility models and Heston-style stochastic volatility models for foreign exchange and equity exposures. A local volatility engine exists in ORE (used for pricing) and is being extended to the exposure simulation framework (in progress), with local volatility for FX now implemented and Equity support work in progress, including a correction for stochastic interest rates. This moves beyond ATM-vol approximations by enabling simulation and pricing that reflects the full implied volatility smile surface. This improves accuracy for path-dependent and smile-sensitive FX and equity positions. In parallel, a Heston model with piecewise-constant parameters is under development (expected in a future release once thoroughly tested). Following these, a stochastic-local volatility (SLV) hybrid model is planned – combining the Heston dynamics with a local volatility surface overlay – to achieve both realistic stochastic behaviour and exact calibration to market option surfaces. These advanced models are on the roadmap (not yet in production as of this writing). Once delivered, they will empower ORE users to match the practices of leading XVA desks that supplement Black-Scholes with local vol or SV models for exotic portfolios. Notably, the addition of SLV will allow replication of observed smiles while retaining a sound stochastic backbone – a capability increasingly sought for managing jumpy market conditions.

2. Commodity Models: Multi-factor Commodity Risk Factors. To better serve commodity XVA use cases, ORE plans to extend its commodity risk factor modelling beyond the current one-factor approaches. A generalized n-factor commodity model is in development, which will include as special cases the well-known 2-factor Schwartz-Andersen model and the 1-factor Schwartz model (thus covering both mean-reverting and dual-factor dynamics). This will effectively replace the legacy single-factor models like Gabillon with a more flexible framework. The work is in progress with expected completion in the coming months. By expanding commodity modelling options, ORE will cater to institutions that require more accurate forward curve dynamics for CVA on commodity derivatives, while still supporting simpler Black-Scholes-style approximations as needed. (As noted earlier, about half of XVA desks use basic Black-Scholes for commodity exposures and the rest use bespoke models – the roadmap will allow ORE to cover both ends of that spectrum.)

3. Collateral and Capital: Margin Simulation and AAD.  ORE integrates the ISDA SIMM framework for initial margin within its American Monte Carlo (AMC) exposure framework, leveraging adjoint algorithmic differentiation (AAD) to compute path-wise sensitivities and margin valuation adjustment (MVA) efficiently. This Dynamic SIMM methodology, based on AMC simulation with fast sensitivity calculation along paths, enables accurate forecasting of future initial margin requirements across simulated market scenarios. The foundation and implementation approach are detailed by Caspers and Lichters ("Dynamic SIMM", October 2025), which demonstrates the method's accuracy against full SIMM recalculations and its superiority over regression-based approaches. Over the longer term, additional regulatory-driven enhancements are being evaluated to ensure ORE keeps pace with emerging best practices. All such extensions will be introduced in a controlled, transparent manner, with clear documentation and testing, so that firms can adopt them confidently when they go live.

Throughout these developments, ORE’s developers clearly distinguish what is live versus upcoming. New model additions undergo rigorous testing and validation (including LSEG’s internal model risk governance for ORE releases) before being officially supported. This transparent evolution process means that users can plan their adoption of new features in line with their own validation timelines. Importantly, ORE’s open-source enhancements flow directly into Risk Analytics Lab as the backend engine. RAL clients therefore automatically benefit from each ORE update as it is released – for example, when the multi-factor Hull-White became available, RAL users could immediately utilize it for XVA runs without any installation or upgrade on their part. Conversely, self-hosted ORE users have the freedom to upgrade at their own pace, staying on a given version until they’re ready to incorporate new models. In both cases, the functional parity between ORE and RAL is maintained. This ensures that whether a client opts for local deployment or the cloud service, they are always leveraging the latest industry-aligned methods supported by ORE.

To complement both deployment options, Post Trade Solutions offers Expert Services supported by a team of quantitative analysts with extensive experience across buy-side institutions, sell-side dealers, and academic research. Customized service-level agreements provide 24/5 technical support, priority bug fixes, bespoke model extensions, and implementation guidance for both local ORE installations and RAL deployments. This professional services layer allows institutions to accelerate their adoption of advanced XVA capabilities while maintaining rigorous internal validation standards. Whether firms choose self-hosted or cloud-based infrastructure, they have access to the same high-caliber expertise and support.

The combination of ORE and Risk Analytics Lab offers a compelling, future-proof solution for XVA analytics that marries industry best practices with deployment flexibility. This approach aligns with the financial services industry's rapidly growing adoption of open-source solutions. According to FINOS's 2025 State of Open-Source report, financial services contributors on GitHub grew 36% from 2021 to 2025, while total commits increased 80% over the same period. Organizations are realising substantial benefits, with nearly half reporting annual cost savings exceeding $500,000 from open-source adoption. ORE's foundation of validated, widely used models gives institutions confidence that their XVA calculations are on the same footing as legacy platforms. At the same time, the ongoing enhancements on the roadmap demonstrate a commitment to meeting emerging needs, such as more sophisticated volatility modelling and integrated margin analytics, in a careful, non-proprietary way. By delivering this capability in dual form – either as a traditional in-house engine or as a cloud-hosted service – ORE and RAL accommodate the full spectrum of operating models, from conservative on-premises setups to cutting-edge cloud adoption.

In a landscape where nearly 70% of XVA desks are embracing cloud computing and virtually all rely on tried-and-true models, ORE and Risk Analytics Lab together serve as a bridge between the old and the new. Firms can modernize their XVA infrastructure and scale their computations on demand – without sacrificing model transparency or alignment with market standards. This balance of innovation and familiarity positions ORE and RAL as strategic tools for institutions aiming to enhance their risk analytics in an efficient, future-aligned manner. ORE and RAL exemplify this evolution toward open, flexible, and cloud-ready risk management infrastructure - offering neutral and authoritative capabilities that are validated by both academic rigor and real-world benchmarks.