With the Fundamental Review of the Trading Book (FRTB) regulation, the Internal Models Approach (IMA) to compute the market risk capital charge changed from a VAR/SVAR calculation to expected shortfall (ES). While these calculations are similar, the new regulation requires up to 63 separate executions, each involving a computationally expensive Monte-Carlo simulation. This constitutes a tremendous computational challenge to banks with IMM approval. It is therefore of paramount importance that the ES implementation is highly optimised.
This paper introduces the calculations required under FRTB-IMA and details several software optimisation techniques to cope with the data and compute complexities involved. Code modernisation techniques as well as accelerator processors are covered (GPU, many-core), recommendations are given, and illustrative examples are shown.