Algorithmic Differentiation: Compliance & Competitive Advantage

Peter Jones Sep 24, 2015

Dr. Russell Goyder

Director of Quantitative Research and Development, FINCAD

Intra Day Risk & Valuation Analytics For Improved Buy-Side Trading Performance

Register & Watch On-Demand Webinar
As calculation complexity and resource requirements scale simultaneously, derivatives dealers are being drawn to algorithmic differentiation (AD) to comply with regulation and pursue competitive advantage.

Interest in AD may be fueled by anticipation of regulations soon to be put into place by the International Swaps and Derivatives Association (ISDA). Starting in September 2016, with full implementation in March 2017, major derivatives firms will be called on to have a transparent and standardized risk-based margining approach for initial margin, called the Standardized Initial Margin Model (SIMM). Part of the ISDA requirement includes capturing first order risk sensitivity with respect to every risk factor on at least a daily basis for determining margin.

As mentioned by ISDA.

“it is imperative that a SIMM model approximates the response to shocks with a fast calculation if we are to avoid derivative price-making coming to a standstill.”
ISDA 2013 Paper - Standard Initial Margin Model for Non-Cleared Derivatives,

With full re-valuations (“bumping”) across a vast and, many times, diverse book of products, major processing strain will certainly be placed on pricing and risk models.

Implementing AD may be a good way for firms to meet the ISDA mandate.

Firms using AD are experiencing an acceleration in risk run-time in the order of 100-1000x that of bumping. Where typical risk runs have been an overnight activity and limited to looking at a subset of total exposure, AD enables firms to view the entire risk landscape on a pre-trade basis. Instead of spending a night to peer into your risk with a flashlight and hoping you looked in the right places, you’re using a floodlight to uncover every exposure – and at a fraction of the cost of bumping.

The only benefit isn’t risk run-time acceleration; while bumping is an approximation, AD gives you exact, analytic Greeks and other sensitivities. As an approximate technique bumping is subject to numerical noise. Among many questions, you’re faced with deciding how big a bump should be, whether a basis point is sufficient and whether it should be relative or absolute.

With AD, this tuning goes away. You get exact, analytic sensitivity. Additionally, when AD is used within calibration, things become more stable. There is no more fine-tuning of risk calculations and your quants are freed up to move onto more productive activities, saving valuable time and resources.

Jasper Andresen, Head of Quantitative Research at Danske Bank was quite right when he said, “The real benefit of AD is not just that I can do things quicker, but the fact that I can start looking at problems I hadn’t dared look at before.” The speed and accuracy of AD are just additional elements of a modern enterprise valuation and risk platform that not only enables regulatory compliance, but also drives competitive advantage.