CLAUDIO ALBANESE, TOUFIK BELLAJ, GUILLAUME GIMONET, GIACOMO PIETRONERO

Quantitative Strategies, Investment Banking Division, Credit Suisse Group, One Cabot Square, London, E14 4Q, UK

Abstract.

Valuing, hedging and securitizing counterparty credit risk involves analyzing large portfolios of netting sets over time horizons spanning decades. Theory dictates that the simulation measure should be coherent, i.e. arbitrage free. It should also be used consistently both to simulate and to value all instruments.This article describes the Mathematics and the software architecture of a risk system that accomplishes this task. The usage pattern is based on an offline phase to calibrate and generate model libraries. Valuation and simulation algorithms are planned offline with portfolio specific optimizations. The interactive user-driven phase includes a coherent global market simulation taking a few minutes and a real time data exploration phase with response time below 10 seconds. Data exploration includes 3-dimensional risk visualization of portfolio loss distributions and sensitivities. It also includes risk resolution capability for outliers from the global portfolio level down to the single instrument level and hedge ratio optimization. The network bottleneck is bypassed by using heterogeneous boards with acceleration. The memory bottleneck is avoided at the algorithmic level by adapting the mathematical framework to revolve around a handful of compute-bound algorithms.