LSM Reloaded - Differentiate xVA on your iPad Mini

46 Pages Posted: 10 May 2017 Last revised: 16 Nov 2018

See all articles by Brian Norsk Huge

Brian Norsk Huge

Saxo Bank

Antoine Savine

Danske Bank; University of Copenhagen - Copenhagen University

Date Written: June 16, 2017


This article by Brian Huge and Antoine Savine reviews the so called least square methodology (LSM) and its application for the valuation and risk of callable exotics and regulatory value adjustments (xVA). We derive valuation algorithms for xVA, both with or without collateral, that are particularly accurate, efficient and practical. These algorithms are based on a reformulation of xVA, designed by Jesper Andreasen and implemented in Danske Bank's award winning systems, that hasn't been previously published in full. We then investigate the matter of risk sensitivities, in the context of Algorithmic Automated Differentiation (AAD). A rather recent addition to the financial mathematics toolbox, AAD is presently generally acknowledged as a vastly superior alternative to the classical estimation of risk sensitivities through finite differences, and the only practical means for the calculation of the large number of sensitivities in the context of xVA. The theory and implementation of AAD, the related check-pointing techniques, and their application to Monte-Carlo simulations are explained in numerous textbooks and articles, including Giles and Glasserman's pioneering Smoking Adjoints. We expose an extension to LSM, and, in particular, we derive an original algorithm that resolves the matters of memory consumption and efficiency in differentiating simulations together with the LSM step.

Keywords: Antoine Savine, Brian Huge, automatic differentiation, Monte-Carlo, financial simulations, risk management, derivatives, LSM, AAD, xVA, check-pointing

Suggested Citation

Huge, Brian Norsk and Savine, Antoine, LSM Reloaded - Differentiate xVA on your iPad Mini (June 16, 2017). Available at SSRN: or

Antoine Savine

Danske Bank ( email )

DK-1092 Copenhagen K


University of Copenhagen - Copenhagen University ( email )

studiestraede 6
copenhagen, 1455


Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
PlumX Metrics