Ahead of the Curve Podcast Series

Join Scott Sobolewski, Roland Stamm, and Joey O’Brien from Post Trade Solutions (formerly Acadia) Quant Services team, as they reflect on a transformative year for the Open Source Risk Engine (ORE). In this episode, they explore major 2024 developments, including enhanced support for PFE and XVA calculations, market risk sensitivity tools, and the growing adoption of ORE across commercial and open-source users. Learn how ORE is powering risk management, model validation, and regulatory compliance for institutions worldwide, and what updates are coming in 2025, from GPU acceleration, hosted analytics services and more.

Hello and welcome to Ahead of the Curve, the podcast from LSEG Post Trade Solutions.My name is Scott Sobolewskipartner in Acadia's Quant services groupand I have with me here today, Doctor Roland Stamm, a fellow partner in Acadia'sQuant services, and Joey O'Brien, a senior consultant on our expert services teamand today we're here to do a year in review for our Open Source Risk Engine, ORE.We would like to highlight some important functionalitythat was recently released over the course of 2024to give you a preview or insight into how clients havemost been using ORE over the course of this yearand what lies ahead in 2025.So if you're a new viewer or listener we'll start first with Joey.What is ORE?Thanks Scott.So ORE is the Open Source Risk Engine.It's a C++ library with over 20 years of developmentto do everything from curve building, pricing of financial instrumentsthe whole way up to complex risk calculations, including XVA’s.It's been used behind the scenes within Acadiafor over five years now in terms of the SIMM calculationso it's been battle tested in that case but it's ongoing and fully open sourcedso that's what we're mainly here to talk about todayand for a bit more background, ORE has been open sourcefor nearly a decade now and has been quite maturewe continue to make continual improvements to ORE to supportclient led growth and other use cases that we see in the open source realm,things like expansion of instrument coverage across all asset classessome other big highlights from this year include SABR modellingfor interest rate options, XVA stress testing and sensitivity calculationinformed by some of our clients, regulatorycapital requirements and also historical simulation barso quite a bit of functionality for derivatives and otherfinancial instruments across market risk and counterparty credit risk.If you're not already aware or you're not familiar with ORE,it is fully open sourced, end to endfor some of our commercial clients who are using ORE in productionor as a replacement for internal or legacyrisk systems, Acadia does also offer commercial supportfor those clients in the form of service level agreements or SLA'sand some of the feedback we've gotten over the last decadesince ORE has been open sourced, is that certainmaybe large banks or other large institutionsare perhaps hesitant to use pure open source software without a commitmentfrom a software vendor.So Acadia has recently solved for that at the beginning of 2024 by offering now,quite a low cost SLA for clients to get aroundsome of those issues in the open source domainbut as you'll find out, plenty of clients,still continue to use ORE entirely for freeso I guess first,I'd say the majority of the development that we've put into our ORE over,the last year in 2024, has been drivenby requests and funded development from some of our commercial clientsso I'll start Roland with you firstmaybe you could cover some of the biggest use cases for our commercial clientsin ORE specifically in the counterparty credit risk space first?Yes so we have one very big client in the USwho has driven a lot of the development that we've done over the past yearand the project that we havethere is really to replace multiple systems in the risk space,both on the market risk side for stress testing and PNL explainbut also on the credit risk sidethere we are talking about XVA calculations and PFE calculationsand maybe at that point Joe, you could divea bit deeper into that PFE piece?So one of the main use cases of ORE withinthe client is in PFE calculationand in some sense PFE calculation is as involvedas it gets from a risk perspectiveit involves simulated market data, repeated pricing and storageso that has been a huge component of the implementation of ORE within that clientin practice, they have a portfolio of over 100,000 tradeswhich at the end of each day they have to calculate PFE’sfor potential futures exposuresin each case and that's a highly involved calculationand we've really battle tested ORE in this case in that implementationto be able to have the capabilityof doing that calculation every single day across that entire portfolioso that's been a one of the primary use cases in the settingand what are the challenges in particularin the potential future exposure simulationso firstly, it generally involves simulating up to 50 years in the futureso you have to be able to simulate market datathat's far into the future along with pricing along those pathsand to get a potential future exposure involves looking at the distributionof potential portfolio valueand that in essence, involves doing that repeatedlymany many times, which makes it computationally expensive.So you run into things like memory issuesfrom a computation perspective, but also storage of the actual resultsso that's been a huge focusthis year is actually making improvements to ORE to be able to do that at scale.Super, thanks.From a use case perspective,the mechanics to simulate, those underlying risk factorsis very similar as it is in PFE to things like XVA and CBAand not only with this client in particular, but we've seen a trendacross the board to unify the modelling, the underlying modelling for XVA and PFEa bit more so in the middle officefor independent risk functions and XVA in the front office roleand Roland maybe you could talk a bit about how this particular client has thought aboutfinding synergies, using ORE to accomplish both XVA and PFE use cases.Yes so the simulations that you would have to doare very similar to what Joey just described that you have to simulatebasically the entire derivatives portfoliothat you are looking at for up to 50 yearsI mean, we have derivatives that run that longso you would have really to go to the maturity of the portfolioand then do multiple paths, usually from a thousandup to 10,000, even 20,000 paths, whatever is requiredIn terms of your computational accuracy that you wantand then even the output is very, very similarthe only difference is that usually peoplethink about XVA in terms of a market priceor an addition to the market price of the portfolio, meaning thatyou calibrate to what's called the risk neutral measure, i.e.you take the volatilitiesthat you find in the market for the underlying instrumentsand on the other hand, for the potential future exposuresimulation people typically use what is called the real world measure, i.e.looking at realised volatilitiesand correlations off the underlying portfolioand what we have found over the course of our consulting businessover the last ten, 15 years is that you can actually arguethat you can use the same calibration for both calculations,and that then offers youa lot of synergies in terms of the computational issues that you run intoYou only have to do this calculation onceyou can reuse the PFE results to calculate XVA’sand that saves you a lot of hardware time,a lot of storage that you would otherwise require and so onso what we see with that client in particular is a hugesaving on an annual basis in the two digit million numberand making that PFE calculation is complicated enough on its ownbut in practice there's an enormous amountand that PFE calculation is what Acadia or what ORE does well, out of the boxbut lots of work in the front end of that processin setting up all of the data layers into ORE for trade data and market dataalso plenty of work on the back end as well, once that calculation is madein the form of model validation and backtesting,and a quick plug for a white paper that these two have authoredtowards the end of 2024 on backtesting risk factors used in PFEsome of those work around the edges that banksor even larger non-banks do for XVA or that’s really where some of those ORE.SLA from our expert services consultants where we can help clientswho are using open source ORE to actually perform the calculationsso counterparty creditrisk has always been a core tenet of what ORE doeswell but it's not exclusively what ORE does.Fundamental to PFEis the ability to make valuations and pricing on the underlying instrumentsbut also market risk is probably the second most common use case.So Roland maybe you can talk tosome of the developmenton the Scripted Trade framework to make market risk sensitivities forthat German client of ours?So Scripted Trade is for me one of the most awesome additionsto the ORE to the open source project because it givesyou really the scope to model any derivativeany pay off that you can think ofbasically, you can havestructured payouts and you can have,termination rights,break rights, make whole clauses, target returnpayments, whatever you can think of, whatever is out there,you can combine those to create reallythe most complex pay offs and pay out structuresso that is really a fantastic thingit is highly competitive nowadays to third party productsthat are out there that you have to pay a lot of money for, rightand one project in particular I think that you were referring to here isone that was just finished very successfullywhere we, replaced a third party productto do pricing and risk management, i.e.market risk sensitivity numbersused both in market risk management but also inhedge accountingusing ORE for highly structured productsboth in the inflation space but also equity and FX linked structuresso that was very very successfuland we are really proud that that worked so wellthe client is extremely happy with the resultsso far a good flavour for how ORE’s been usedcommercially by our clients in production in the front officehedging XVA’s middle office PFEalso what we would traditionally say back officefunctions like independent model risk groups or model validation groupsORE provides a fantastic out-of-the-box capability,whether the model governance or model risk managementtask is validating pricing modelsalso risk models that functionality is available out of the boxand Joey, maybe you could talk a bit to howusing ORE as a benchmark tool on the model validation sidehelps saves save clients enormous amount of, time and effortavoiding the need to recreate benchmark models, from scratch. For sure Scott.So regardless of jurisdiction, I think we can all agree thatevery regulator encourages an independent valuationas part of any good model validation or model governance process wherebyyou have your primary model and you want to have some insuritythat the PFE’s it's producing are actually accurateso there are a number of ways of doing that.some are internal model validation units within a bank,actually implementing things like spreadsheets or their own risk library.Python notebooks. I see all the time. Yep, yepand using those to perform that exercise.and that might just be as part of good practice doing thatperiodically within the bankand in that case, we are seeing a number of clientswho want to actually replace these notebooks, Jupyter notebooks and,spreadsheets with ORE to make it moreboth transparent andusing a library which rather than just being coded up for the day off,is actually battle hardened,with a library that's used by, by the entire industry, every single dayso that's a huge use case for ORE is to do thatkind of benchmarking exercise to perform good model hygiene.Yeah and that's just day to day basiswhat we also see a lot of time is that clientsmaybe have a new regulatory submission.Right.They have a capital model or a pricing modelwhich they want to feed to the regulators.That might be the SEC or the Bank of Englandand as part of that submission process, they will be requiredto do an independent evaluationso throughout the year,we've had a number of projects where we've actually supported clientsin doing that exact exercise using ORE versus their in-house pricing libraryand to my eye anyway and my experience over the course of 2024,an emerging number of buy side non-bank institutionseither subject to increased regulatory scrutiny to stand up a modelgovernance or model risk managementframework for the first time, or they're so large thattheir management, their board has required that those teams startto dip their toes into the model validation model documentation spaceand if that's a topic that's near and dear to your heart, another referenceI'd encourage you to check out a previous podcastthat Roland Stam and I had authored specifically to initial marginmodel validation and model governance from, from late 2024here on the Ahead of the Curve podcastso we're certainlyaware of the commercial use cases for ORE becauseclients are coming to us and telling us about their desire to use itwe are occasional aware about pure open source use casesand I'd say one of the, biggestuse cases so far Acadia, is most well known, I would say for its supportof the uncleared margin rules helping clients make their non clearinitial margin calculations and performing all of the required governancearound thatwe do have several clients who are using ORE purely to calculateSIMM based on locally sourced or locally produced CRIFsand that continues to be a very common use caseAcadia is quite diligent in promoting and pushing outnewer SIMM versions as published by ISDA, making sure they find their wayboth into our hosted commercial services as well as into ourour open source already offeringand even if ORE has not had an official release,we make sure to have off cycle releases to support those newer SIMM versionsso if you are using ORE for SIMMyou can rest well knowing that new versions will be thereand if you need, greater certainty contractually around that, you can,inquire about an SLA to support those local SIMM calculations in ORE.A second interesting use case has been forbalance sheet risk management, let's say, or even pure opensource use cases on the XVA or counterparty credit risk spaceSo Roland, maybe you can talk to some of our experiencesin that department or knowledge of clients using it for those use cases.Sure.So with our help usually,so that's why we know these clients,we have helped them migrate their portfolios to OREand one particular case that you mentioned first is, really theentire balance sheet of a small-ish bankthat has been transferred to OREincluding accounts including money markettrades, bonds,what have you, and derivatives, of courseand to use ORE then to risk managethis portfolio or the entire balance sheet to the pointwhere they can steer their or risk manage their IFRS marginso that is the only case that we are aware of thatan organisation has really put the entire balance sheet into OREbecause most of our users would use ORE only for derivativeswe have some other cases where certain specialinstruments are also priced with ORE like special loansbut this particular client is the only one that we are aware of that is,doing the entire balance sheet within ORE,right?Yeah and XVA and things like CSA evaluation still continue to be a very commonuse case as we mentioned earlier, absolute counterparty credit riskbut you know wewe, I say we as our Quant group here at Acadia,we've been doing XVA modelling since the very early days of OREand that counterparty credit risk stripe within ORE isI would say the most battle tested here over the last 15 plus years. Yes.That's kind of where it all started.Really. Yeah, actually.and those listeners who have been followingalong Ahead of the Curve for the last few years know that,the majority of us here in the Quant services groupare recent joiners to the LSEG family.Acadia having been acquired just about two years agofully by, the LSEG Post Trade Solutions Group.Over those last two years, we've found, several opportunities hereinternally within LSEG particularly within Post Trade Solutions,where LSEG itself has been reliant on third party risk and pricing vendors.and we've identified a number of opportunities internally to,replace some of those vendors withORE as well with LSEG essentially becoming a client of ORE.So Joey, I was hoping you could expand a bit on the particularuse case for a sister company of ours internally here within PTS Quantile and some of their use cases for ORE.Sure Scott. So within the LSEG universe, of course,there's a huge number of groupsthat are focussed on similar topicsand one of them was Quantile.They're focussed in the compression space but to do the compressionthey have to be able to calculate SIMM repeatedly essentiallyand that involves pricing.So with that in mind, they had a licensed platformthat they were usingand paying a hefty license fee to do so in order to perform those calculationsbut as part of Acadia's move into this universewe saw opportunities where maybe they could actually use ORE to doa lot of the feeder calculations that they need to use their service.So things like curve building and pricing.So over the past year or so, there's been a huge focus on actuallyimplementing ORE within other groups such as Quantile within LSEGand leveraging that to replace these legacy vendor systemsover the course of a year and it's something we hope to continueto do over the coming years as well.Yeah, we've been able to declare success on that Quantile transitionat the end of 2024, and hopefully more to comeon other business units within LSEGI think it's particularly very excitingthat LSEG has not only continued to promotethe sponsorship of the open source risk project,it's actually accelerated the ability to participate in the open source realm.Commercially and business wise is also throwing its full weight behindadoption of ORE internally.So as we near the conclusion here,I know the majority of this was a look back into how ORE was used in 2024but let's spend a few minutes looking ahead to 2025 and beyond.We do have an upcoming ORE release this spring 2025and I'd say the majority ofwhat you'll see in those release notes here, upcoming over the next few monthsare driven by some of those commercial clients that we mentioned earlier in the podcastbut in the meantime, there's also one exciting initiativeAcadia is moving forward with on a commercial basis to expose OREmore easily to clients to allowour clients and other users to demo OREwithout fully implementing it locally.So avoiding some of those headaches around market data configuration.Again, another topic that myself and James Mac Hale had talked aboutin a separate podcast, again late last year, but also having all of the requiredconfigurations preset and customised to those input portfolios.So if you're familiar with Acadia, we have worked very hard over the last5 to 7 years to build out a centralised market data servicethat we use in our hosted commercial initial margin services,and we're using that same market data service nowto power ORE on a hosted basis through Jupyter notebooksfor a suite of analytics that expand well beyond initial margin.So I'm quite excited by the potential opportunity fora much larger number of people to get familiar with the capabilities of OREwithout having to jump over as many hurdles as they havein 2024 and in previous years.But in terms of specific development in 2025Roland, what can some of our ORE users or interested partiesexpect in that upcoming spring release later this year?So we will be building out ORE as a service, meaning that so far in the past ORE has beenwhat's called a stateless engine.You called it, it did its thing and it shut down againand what we have been working onfor one of those clients that we mentioned beforeis that ORE is now capable of staying in a certain state,staying alive, waiting for new input.Waiting for updates let's say on market dataand then recalculate stuffwithout shutting downso that you save time,which is especially importantfor those performance heavy calculations that we mentioned before.Let's say you want to do just a marginalPFE calculation for a new trade or something like thatand you already have all the PFE calculationsfor your old portfolio from yesterday let's say.So, these kinds of things.and in that vein, we will have,more services to come to what we have already published,for example, the LGM calibration will become a service, i.e. acalibration that you can write outand then reuse, reload into ORE.Yeah, that would be one thing.As the instrument coverage and the analytics coverage has expandedwithin ORE over the last, close to a decade now, I'd say a greaterfocus from some of our commercial users, particularly those with largerportfolios, has been on performance optimisation and performance improvements.So in previous ORE releases, we've supported multithreading.I mentioned XVA sensitivities, which the XVA calculation in itself is quite burdensome.XVA sensitivities is an order of magnitude more difficult,you know, that has been released.The ability to make a basic XVA sensitivity calculationwas was released in 2024but I know, Roland, we've expanded the performance ofthose XVA sensitivities using algorithmic differentiationand that's an exciting release that should be coming to this spring as welland also, the usage of GPUs to speed up things is also being expanded on.There was a proof of concept release, so to say, in the last release.But, we are expanding on that and hopeto see even better results and broader coverage.We will also. Sorry for interrupting.We will also include more asset classes in the American Monte Carlo framework,especially equities and probably also commodities.I'd say the biggest benefit of using ORE or familiarising yourself with OREis as you've hopefully learned over the course of this podcast,the amount of commercial clients and open source usersthat are feeding back development into this life cycle,it has as, we know for a factwe use ORE ourselves here in our hosted initial margin services.Over 100 clients use it every day to make those initial margin calculations.So not only, consulting clients of ours that we know about,sort of either funding or contributing directly to OREbut also development that we make ourselvesto service our our hosted risk services clientsand the GPU topic sparked my memory there,Joey, maybe we could get a bit into some of the specific development that,clients will see upcoming here in 2025related to enhancements that we've made ourselvesto support those risk services clientsGPUs and backtesting is one one area that comes to mind.For sure so these extremely computationally heavy exercises such as backtesting,whereby you might have to calculate sensitivities or PNL’sover a 10 or 15 year period, that's a performance heavy task.So we're always looking for tricks we can do to speed up that calculation.One of them is a joint algorithmic differentiation,which I think is a really cutting edge topicin terms of performance in that space.So we have an implementation of that within OREwhich makes that kind of scale of back testingfor a number of clients possible on a frequent basis.So that's probably the first major performance enhancementthat we've seen over this yearand then the other ones, like you both mentioned before, is the GPU spacewhich allows a lot of enhancements to performance calculations being madeso the industry knows, even beyond the industry knows that GPUs are pretty hot right now.for a range of intensive calculations.So we also have ideas of leveraging these GPUsand maybe a framework in order to enhance our abilityto do these kind of CVA sensitivity calculationsthat you mentioned earlier, Scott, that are extremely computationally heavy.I think one CVA calculation is hard at times,but doing it to calculate sensitivities for every risk factor,you do need these kind of performance enhancements.and we've embraced them fully in order to be able to do thatand we've been working on this GPU research and testing for 2 to 3 years nowto the point where we're excited to commit to releasingORE’s compatibility with GPU frameworks OpenCL and Cudain the open source domainupcoming this spring 2025so I think that's certainly something to to look forward toand I would say the last thingrelated to our hosted commercial services, Acadia does offer acapital calculation service alongside our daily initial margin services,whereby some of the standardised Basel three capital metrics can becalculated by Acadia now using using OREso things like SACCR that your standardised approachto counterparty credit risk under Basel three, your FRTB’s standardised approachvery similar to the SIMM framework if you're familiar with that as a sensitivity based Far approximationand I'd say more cutting edgeor more exciting, not only BI CVA but also SI CVA,we're planning to open source all of thoseBasel three standardised metrics here upcoming this spring within ORE.So we're working very hard now, not only onensuring the accuracy of those analytics, but also ensuring thatthe instrument coverage associated with those capital metrics isas close to what you'll see across the entire ORE suite.So I think that's, near and dear to my heart coming from the US, where,Basel III endgame has been quite the hot topic here over the last year or so.So with that, I'd like to thank both Roland and Joey for joining me todayon this exciting, I hope, episode of Ahead of the Curve.For more information about the Open Source Risk Engine,I encourage you to visit our website, opensourcerisk.organd if you're interested in any of the commercial servicesthat you heard us mentioned today, in particular the service levelagreement or any sort of commercial supportthat you may require for the Open Source Risk Engine,you can visit acadia.inc to get in touch with some of our sales team.So I appreciate it.Thank you guys, and have a great rest of your day.Thank you, you too.Thanks.

Ahead of the Curve provides insightful perspectives and insights into the margin and collateral industry. Hosts are joined by special guest speakers from across the industry to share topical perspectives, as they aim to really get ‘under the skin’ of the issues that are transforming the sector.

Centralise

Providing a golden source of trade and agreement data and standardising portfolios on central open-access platforms, including SwapAgent.

Streamline

Discover industry-leading solutions to streamline your portfolio, including IM Exposure Manager, Collateral Manager and Margin Manager.

Optimise

Unlocking greater efficiencies and reducing counterparty risk through portfolio optimisation.

Analyse

Access risk and data tools to aid regulatory compliance and gain a full, real-time view of your trade exposures.

Advise

Working in partnership with market participants, we’re ensuring the market evolves with stability, security and innovation at its core.

Our compression and optimisation services are delivered by Quantile. View Quantile regulatory information.

Ahead of the Curve podcast

2024 Year in Review – Open Source Risk Engine (ORE)

Listen to the podcast

Also available on

Previous episodes

Navigating Counterparty Credit Risk & NBFI Oversight

Market Outlook 2025 - A focus on the OTC Derivatives landscape

ISDA SIMM 2.7: Challenges and opportunities